Methods
in
Molecular Biology™
Series Editor John M. Walker School of Life Sciences University of Hertfordshire Hatfield, Hertfordshire, AL10 9AB, UK
For other titles published in this series, go to www.springer.com/series/7651
Systems Biology in Drug Discovery and Development Methods and Protocols
Edited by
Qing Yan PharmTao, Santa Clara, CA, USA
Editor Qing Yan PharmTao Santa Clara, CA USA
[email protected]
ISSN 1064-3745 e-ISSN 1940-6029 ISBN 978-1-60761-799-0 e-ISBN 978-1-60761-800-3 DOI 10.1007/978-1-60761-800-3 Springer New York Dordrecht Heidelberg London Library of Congress Control Number: 2010934107 © Springer Science+Business Media, LLC 2010 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Humana Press, c/o Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. While the advice and information in this book are believed to be true and accurate at the date of going to press, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Cover illustration: See Figure 1 of Chapter 13 for more information. Printed on acid-free paper Humana Press is part of Springer Science+Business Media (www.springer.com)
Preface Currently, the drug discovery industry has reached the bottleneck. Adverse drug reactions (ADRs) are one of the leading causes of death and illness in the US. In the mean time, although the industry is spending a tremendous amount of money and time, high-profile drug withdrawals are increasing, with fewer FDA approvals of new drugs. The “one-drugfits-all” model has not been successful. There is an urgent need to change the current drug discovery and development process that has high cost, low efficacy, and high ADRs. It is necessary to develop personalized medicine that treats whole systems and brings the right drug to the right patient with the right dosages. Systems biology emerged with the realization that genes, molecules, tissues, and organs do not work alone but interact with each other in a whole system. Combined with pharmacogenomics studies, systems biology would provide a holistic and thorough understanding of health and medicine. Such understanding would change the emphasis of medicine from diseases to humans and enable the transformation from disease treatment to prevention and health promotion. This book has several features that readers may find helpful to their work. First of all, it focuses on translational methods through applying systems biology approaches directly in drug development and clinical practice. One of the major challenges that needs to be resolved in current bioscience is the translation of basic studies into better clinical outcomes. This book is written in response to this challenge through highlighting the development of translational medicine based on systems biology. We hope that these approaches may help make some breakthroughs and advancement toward the realization of personalized medicine, which is also the second feature of the book. That is, most of the methods and protocols described in the book are geared toward the development of individualized therapeutics. The third feature is that this book provides both practical methods and comprehensive resources that can be used for solving complex problems in medicine. A wide range of approaches are introduced with problem-solving objectives, from theoretical and computational analyses to experimental steps. The fourth feature is that this book integrates the advancement of science with innovative technologies. While the first part of the book describes cutting-edge technologies and methods in the field, the second part illustrates how the technologies can be applied in science for disease understanding and therapeutic discovery. The first part of the book introduces basic and novel concepts, as well as advanced technologies in systems biology for efficient drug discovery and development. Such concepts include proteomics, cell behavior, interactomes, and multi-drug targets. The technologies include computational modeling, Bayesian networks, translational bioinformatics, quantitative proteomics methods, microarrays, and RNA interference (RNAi). These technologies can help us with the identification of biomarker genes and pathways and understanding the interactions among genes, drugs, and diseases. One of the potential results from systems biology studies is the identification of novel drugs tailored for individuals. Concepts such as proteomics, toxicoproteomics,
v
vi
Preface
epigenetics, and their roles in systematic drug target discovery and clinical trial design are introduced in the first chapter of this book. For instance, current proteomics technologies include two-dimensional gel electrophoresis (2D-GE), mass spectrometry (MS), and protein arrays. Based on proteomic studies, multiple-target approaches are novel ways for the design of drugs against atherosclerosis, cancer, depression, psychosis, and neurodegenerative diseases (see Chapter 2). Novel computational and mathematical modeling are the essential methods for dealing with complex proteomic data and understanding of genetic interaction networks involved in these processes. Quantification approaches are important for the identification of protein biomarker signatures and the study of interactomes. For example, stable isotope labeling by amino acids in cell culture (SILAC) is a quantitative proteomics method (see Chapter 3). It can be combined with high-resolution MS as a potent tool for functional analyses. In combination with RNAi, SILAC can address many of the systems-wide approaches that were previously impossible. Better understanding of intracellular and cellular circuits would contribute to systems biology approaches to drug discovery. Mathematical modeling of molecular modules in a cell can link intracellular molecular machinery and cellular activity to enable the understanding of cell behavior. The cell behavior includes avoidance reaction, escape reaction, conjugation, chemotaxis, cell division, stochastic ball movement, search reaction, and Ca2+ concentration-dependent movement (see Chapter 4). The elucidation of regulatory networks and pathways from proteomic data reveals how proteins regulate each other, which is important for drug design. Computational methods to the understanding of the functional roles of cellular networks include “static” models, as well as dynamical and stochastic simulations (see Chapter 5). These methods are useful for interpretation of high-throughput interaction data, finding gene expression patterns, and building predictive models. For instance, Gaussian Bayesian network methodology is useful for the analyses of static and dynamic time series data (see Chapter 6). Bayesian network inference methods in the analysis of flow cytometry data can be used to evaluate regulatory network topology. Furthermore, analyses of the large-scale self-regulatory behavior of the cell may help establish comprehensive models of the cell and genes action (see Chapter 7). Such methods for analyzing microarray data include principal component analysis (PCA), clustering, tree building, self-organizing map (SOM), and bootstrapping. Based on these data mining approaches, translational bioinformatics is a powerful method to bridge the gap between systems biology research and clinical practice. Translational bioinformatics would bring novel insights in the identification of biomarkers and systemic interactions. Methods of data integration and data mining can provide decision support for both researchers and clinicians (see Chapter 8). Part II of this book focuses on the application of these methods in the translation of systems biology into understanding of disease states and development of personalized therapeutics. These diseases include cardiovascular disease, cancer, lupus erythematosus, influenza, drug abuse, and brain injury. Most of these diseases have close associations with certain responses such as inflammation and immunological reactions. For example, inflammation is a complex response involved in many diseases including rheumatoid arthritis, asthma, cancer, diabetes, atherosclerosis, Alzheimer’s, and obesity. Translational applications of computational simulations applied to inflammation are reviewed in this book, such as agent-based modeling (ABM) and equation-based modeling (EBM) (see Chapter 9). Translational systems biology modeling efforts at various levels, from the systemic level to the cellular level are described.
Preface
vii
The immune system plays an important role in the development of personalized medicine for a variety of diseases including cancer, autoimmune diseases, and infectious diseases. The integration of immunoinformatics with systems biology may lead to a better understanding of immune-related diseases at various levels for the development of individualized vaccines and drugs. Basic concepts and various methods are explained in detail with examples in different diseases (see Chapter 10). In the development of cardiovascular disorders such as atherosclerosis, chronic inflammation also plays an important role. Systems biology tools and comprehensive bioinformatics packages can be used for the development of cardiovascular drugs and the detection of beneficial and adverse effects. Detailed protocols and methods are described in this book, from experimental design and tissue collection, to bioinformatics analysis of functional processes and pathways (see Chapter 11). Cancer is a complex disease at tissue, organism, and population levels. Various computational and mathematical approaches are necessary for analyzing different types of cancer data. Comprehensive approaches are discussed in this book, from data- and process-driven theoretical methods to experimental methodologies in cancer research, as well as applications in the clinical context (see Chapter 12). Systemic lupus erythematosus (SLE) is a disease with inappropriate response to selfantigens. Different levels of mechanisms and causes are reviewed in this book, including genetic, environmental, and hormonal factors (see Chapter 13). Influenza virus infection is a public health threat worldwide. It is urgent to develop effective methods and tools for the prevention and treatment of influenza. New personalized vaccines, adjuvants, and drugs may result from the understanding of interactions of host genetic, environmental, and other factors. A comprehensive overview is provided in this book about systems biology studies for the development of the optimal prevention and treatment strategies for influenza (see Chapter 14). Drug abuse, such as the abusive use of methamphetamine (METH), is a growing problem that may cause harmful effects to the human brain. This book describes the current protocols for analyzing the drug abuse problem, from the in vitro cell culture to in vivo rat models, as well as bioinformatics methods for establishing interaction maps to study altered functions (see Chapter 15). Traumatic brain injury affects millions of patients every year in the USA, but currently no FDA-approved drugs are available to treat the problem. Traditional single drug-totarget approach has failed to work. Systems biology-based approaches may provide promising solutions through focusing on converging pathways and downstream biomarkers as potential therapeutic intervention and targeting points (see Chapter 16). By covering topics from fundamental concepts to advanced technologies, this book can be used by biomedical students and professionals at all levels who are interested in integrative studies in molecular biology, genetics, bioinformatics, bioengineering, biochemistry, physiology, pathology, microbiology, immunology, pharmacology, toxicology, drug discovery, and clinical medicine. Written by leading experts in the field, this book intends to provide a practical, state-of-the-art, and holistic view of the translation of systems biology into better drug discovery and personalized medical practice. I would like to thank all of the authors for their innovative input and sophisticated contributions to this exciting new field. I also thank the series editor, Dr. John Walker, for his help with the editing.
Qing Yan, MD, PhD Santa Clara, CA
Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
v xi
Part I Methodologies of Systems Biology in Drug Discovery and Development 1 Proteomics and Systems Biology: Application in Drug Discovery and Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Bulbul Chakravarti, Buddhadeb Mallik, and Deb N. Chakravarti 2 Systems Biology Approaches and Tools for Analysis of Interactomes and Multi-target Drugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 André Schrattenholz, Karlfried Groebe, and Vukic Soskic 3 Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics and RNAi Approach in Drosophila melanogaster . . . . . . . . . . . . . . . . . 59 Alessandro Cuomo and Tiziana Bonaldi 4 Systems Biology of Cell Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 Najl V. Valeyev, Declan G. Bates, Yoshinori Umezawa, Antonina N. Gizatullina, and Nikolay V. Kotov 5 Computational Modeling in Systems Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 Ravishankar R. Vallabhajosyula and Alpan Raval 6 An Introduction to Gaussian Bayesian Networks . . . . . . . . . . . . . . . . . . . . . . . . . 121 Marco Grzegorczyk 7 Derivation of Large-Scale Cellular Regulatory Networks from Biological Time Series Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 Benjamin L. de Bivort 8 Translational Bioinformatics and Systems Biology Approaches for Personalized Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Qing Yan
Part II Systems Biology Methods for Disease Treatment and Translational Medicine 9 Systems Biology and Inflammation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 Yoram Vodovotz and Gary An 10 Immunoinformatics and Systems Biology Methods for Personalized Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 Qing Yan 11 Systems Biology Approaches to the Study of Cardiovascular Drugs . . . . . . . . . . . 221 Yuri Nikolsky and Robert Kleemann
ix
x
Contents
12 Cancer Systems Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dana Faratian, James L. Bown, V. Anne Smith, Simon P. Langdon, and David J. Harrison 13 Systemic Lupus Erythematosus: From Genes to Organ Damage . . . . . . . . . . . . . . Vasileios C. Kyttaris 14 Systems Biology of Influenza: Understanding Multidimensional Interactions for Personalized Prevention and Treatment . . . . . . . . . . . . . . . . . . . . Qing Yan 15 Methods in Systems Biology of Experimental Methamphetamine Drug Abuse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Firas H. Kobeissy, Shankar Sadasivan, Melinda Buchanan, Zhiqun Zhang, Mark S. Gold, and Kevin K.W. Wang 16 Systems Biology and Theranostic Approach to Drug Discovery and Development to Treat Traumatic Brain Injury . . . . . . . . . . . . . . . . . . . . . . . . Zhiqun Zhang, Stephen F. Larner, Firas Kobeissy, Ronald L. Hayes, and Kevin K.W. Wang
245
265
285
303
317
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
Contributors Gary An • Department of Surgery, Center for Inflammation and Regenerative Modeling, Northwestern University, Chicago, IL, USA; McGowan Institute of Regenerative Medicine, University of Pittsburgh, Pittsburgh, PA, USA Declan G. Bates • Systems Biology Lab, Department of Engineering, University of Leicester, Leicester, UK Tiziana Bonaldi • Department of Molecular Oncology, European Institute of Oncology, Milano, Italy James L. Bown • University of Edinburgh, Edinburgh, UK Melinda Buchanan • Department of Psychiatry, Center for Neuroproteomics and Biomarkers Research, McKnight Brain Institute of the University of Florida, Gainesville, FL, USA Bulbul Chakravarti • Department of Biology and Department of Chemistry, York College, City University of New York, CA, USA Deb N. Chakravarti • Department of Biology and Department of Chemistry, York College, City University of New York, CA, USA Alessandro Cuomo • Department of Molecular Oncology, European Institute of Oncology, Milano, Italy Benjamin L.de Bivort • Rowland Institute at Harvard, Cambridge, MA, USA Dana Faratian • Breakthrough Research Unit, Centre for Research in Informatics and Systems Pathology (CRISP), University of Edinburgh, Edinburgh, UK Antonina N. Gizatullina • Biophysics & Bionics Lab, Department of Physics, Kazan State University, Kazan, Russia Mark S. Gold • Department of Psychiatry, Department of Community Health and Family Medicine, Center for Neuroproteomics and Biomarkers Research, McKnight Brain Institute of the University of Florida, Gainesville, FL, USA Karlfried Groebe • ProteoSys AG, Mainz, Germany Marco Grzegorczyk • Department of Statistics, TU Dortmund University, Dortmund, Germany David J. Harrison • University of St. Andrews, St. Andrews, UK Ronald L. Hayes • Banyan Biomarkers, Inc., Alachua, FL, USA Robert Kleemann • BioSciences, Department of Vascular and Metabolic Diseases, TNO-Quality of Life, Leiden, The Netherlands Firas H. Kobeissy • Department of Psychiatry, Center for Neuroproteomics and Biomarkers Research, McKnight Brain Institute of the University of Florida, Gainesville, FL, USA Nikolay V. Kotov • Biophysics & Bionics Lab, Department of Physics, Kazan State University, Kazan, Russia Vasileios C. Kyttaris • Division of Rheumatology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA
xi
xii
Contributors
Simon P. Langdon • University of St. Andrews, St. Andrews, UK Stephen F. Larner • Banyan Biomarkers, Inc., Alachua, FL, USA Buddhadeb Mallik • Keck Graduate Institute of Applied Life Sciences, Claremont, CA, USA Yuri Nikolsky • GeneGo, Inc., St.Joseph, MI, USA Alpan Raval • Keck Graduate Institute of Applied Life Sciences, Claremont, CA, USA; School of Mathematical Sciences, Claremont Graduate University, Claremont, CA, USA Shankar Sadasivan • Department of Psychiatry, Center for Neuroproteomics and Biomarkers Research, McKnight Brain Institute of the University of Florida, Gainesville, FL, USA André Schrattenholz • ProteoSys AG, Mainz, Germany V. Anne Smith • University of Abertay Dundee, Dundee, UK Vukic Soskic • ProteoSys AG, Mainz, Germany Yoshinori Umezawa • Kings College London, St John’s Institute of Dermatology, London, UK Najl V. Valeyev • Systems Biology Lab, Department of Engineering, University of Leicester, Leicester, UK; Kings College London, St John’s Institute of Dermatology, London, UK Ravishankar R. Vallabhajosyula • Keck Graduate Institute of Applied Life Sciences, Claremont, CA, USA Yoram Vodovotz • Center for Inflammation and Regenerative Modeling, McGowan Institute of Regenerative Medicine and Department of Surgery, University of Pittsburgh, Pittsburgh, PA, USA Kevin K.W. Wang • Department of Psychiatry, Center for Neuroproteomics and Biomarkers Research, McKnight Brain Institute of the University of Florida, Gainesville, FL, USA; Center of Innovative Research, Banyan Biomarkers, Inc., Alachua, FL, USA Qing Yan • PharmTao, Santa Clara, CA, USA Zhiqun Zhang • Center of Innovative Research, Banyan Biomarkers, Inc., Alachua, FL, USA
Part I Methodologies of Systems Biology in Drug Discovery and Development
Chapter 1 Proteomics and Systems Biology: Application in Drug Discovery and Development Bulbul Chakravarti, Buddhadeb Mallik, and Deb N. Chakravarti Abstract Studies of complex biological systems aimed at understanding their functions at a global level are the goals of systems biology. Proteomics, generally regarded as the comprehensive study of the expression of all the proteins at a particular time in different organs, tissues, and cell types is a key enabling technology for the systems biology approach. Rapid advances in this regard have been made following the success of the human genome project as well as those of various animals and microorganisms. Possibly, one of the most promising outcomes from studies on the human genome and proteome is the identification of potential new drugs for the treatment of different diseases and tailoring the drugs for individualized patient therapy. Following the identification of a new drug candidate, knowledge on organ and systemlevel responses helps prioritize the drug targets and design clinical trials based on their efficacy and safety. Toxicoproteomics is playing an important role in that respect. In essence, over the past decade, proteomics has played a major role in drug discovery and development. In this review article, we explain systems biology, discuss the current proteomic technologies, and highlight some important applications of proteomics and systems biology approaches in drug discovery and development. Key words: Proteomics, Systems biology, Drug development, Drug discovery, Mass spectrometry, Two-dimensional gel electrophoresis, 2D-GE, Protein arrays
1. Introduction Systems biology refers to the study of the interrelationships of all the different elements in a biological system and not studying them in an isolated manner one at a time. As explained by Weston and Hood (1), biological information moves from the genome of organisms to ecologies, the relationship between organism and their environment, in a hierarchical manner (DNA → RNA → protein → biomodules, protein interactions, protein or gene regulatory networks → cells → organs → individuals → populations
Qing Yan (ed.), Systems Biology in Drug Discovery and Development: Methods and Protocols, Methods in Molecular Biology, vol. 662, DOI 10.1007/978-1-60761-800-3_1, © Springer Science+Business Media, LLC 2010
3
4
Chakravarti, Mallik, and Chakravarti
of individuals → ecologies). As evident from the above hierarchy, changes in environmental signals can change the biological information at each level. Thus, for the systems biology approach, it is important to gather information at as many levels as possible followed by the integration and meaningful interpretation of such data in the context of the biology of the specific organism. By integrating the biological knowledge, the ultimate goal of systems biology is to understand how the molecules act together within the network of interactions that makes up life. One of the major challenges of systems biology is to determine the architecture of protein and gene regulatory networks and to understand how their behaviors are integrated to carry out biological functions. Progress in systems biology research has been possible due to advancements as well as cross-disciplinary research among different branches of science and technology such as biology, chemistry, computer science, bioinformatics, engineering, mathematics, and physics. In this regard, specific mention should be made of the human genome project describing the human genes and ciscontrol elements; the availability of high throughput platforms for genomics, proteomics, and metabolomics, which makes possible the rapid acquisition of global data sets; high speed DNA sequencers; DNA arrays; rapid methods for genotyping; high throughput proteomics capability including protein chip array and particularly mass spectrometry (MS); availability of the internet for acquiring and disseminating the knowledge-base of large global data sets on DNA, RNA, proteins as well as their interactions. For diagnosis as well as prognosis of any disease, it is fundamentally important to understand the underlying molecular basis associated with the disease. Until very recently, the detection of diagnostic biomarkers as well as the development of a drug target was focused mostly on single molecules. Due to rapid technical advancement particularly in omics biology, such as genomics, proteomics, and metabolomics, it is becoming increasingly evident that a cellular behavior associated with normal physiology or a disease pathology is the outcome of interaction at various levels which take place among different cellular components. One of the major goals of systems biology is to understand the role of protein biomodules (groups of proteins that perform a particular function such as galactose metabolism, protein synthesis, etc.) as well as their interconnections which give rise to networks in physiology and pathology. It is thus important to conduct basic research on a systems biology-based approach to understand the normal biological systems as well as the pathological states. The successful treatment of many diseases will depend on overcoming genetic or protein defects such as genetic mutation, aberrant protein– protein interaction, protein misfolding, protein mislocalization, etc.
Proteomics and Systems Biology: Application in Drug Discovery and Development
5
Knowledge of the physiologically healthy system will help us pinpoint any such defect present in a diseased system. With the help of continuously growing technology platforms, particularly due to rapid advancement of the proteomic technology, it is becoming increasingly feasible to carry out a systems biology approach to profile global cellular and molecular changes associated with healthy and diseased state. This level of information is helpful for a more rational and effective drug design to overcome the malfunctioning of the network and come up with a better treatment strategy.
2. Proteomics and Systems Biology Proteome includes all proteins present in a cell, tissue, or organ at a specific point of time. Analogous to the term genomics, which involves the studies of an organism’s entire genome, the studies on the entire set of proteins is called proteomics. Proteomics, the parallel separation, detection, quantification, and identification of all proteins present in a cell, tissue, or organ as well as the analysis of their properties such as posttranslational modifications and interactions provide a more detailed information about any particular biological system compared to that obtained from the genome or mRNA expression profiling of that system. A vast amount of DNA sequence information is available for both eukaryotes and prokaryotes whose entire genome has been sequenced. The genome is rather static and mRNA profiling gives a snapshot of gene expression level at a certain time and condition. Although nucleotide microarray for mRNA profiling is useful for multiplexed comparative analysis of gene expression levels which provide important insights into various cellular mechanisms under different conditions (2), proteins are in general far more directly involved in performing almost all of the cellular activities. Interestingly, as it is turning out, the number of proteins in a cell is much higher than the number of genes due to differential splicing and different posttranslational modifications. Apart from this, sometimes there is a lack of correlation between mRNA expression and protein expression levels due to varying stabilities and life times (3–5). In fact, rather than just providing quantitative measurement of different proteins, detailed proteomic analysis provides lot more information such as posttranslational modification, subcellular localization, etc. As a result, the analysis of the proteome is far more informative than mRNA expression profiling and offers information with different type of value. Proteomics is becoming increasingly important to enhance our knowledge in biological as well as medical research. However, studies on protein expression face several analytical challenges since the proteins are
6
Chakravarti, Mallik, and Chakravarti
expressed over a wide dynamic range, 1–106 in cells, and greater than 1–109 in serum. While the most abundant proteins in serum such as albumin is present in mg/mL level (30–50 mg/mL), low abundant proteins such as IL-6 is present in pg/mL level (0–5 pg/mL) (reviewed by Anderson and Anderson (6)). However, no single method can measure protein expression levels over such a wide dynamic range. Also, there is no amplification method for proteins analogous to the polymerase chain reaction (PCR) for amplification of genes. In addition, there is a change in protein expression pattern under various developmental, physiological, pathological, and environmental conditions which requires a way of taking global snapshots of patterns of protein expression. The combination of different proteomic technologies is required for such purpose. The goal of proteomics is multifaceted and involves a study of proteins with reference to their expression, posttranslational modification, protein–protein interaction, protein–DNA interaction, protein–lipid interaction, protein processing and turn over, cellular and subcellular localization, at a global level. In fact, significant progress has been made toward achieving these goals since the term “Proteomics” was coined (7). However, it is necessary to integrate proteomic data with other information such as gene, mRNA, and metabolite profiles to fully understand how the system works.
3. Current Proteomic Technologies
As explained above, there is currently a vast collection of DNA sequence information as well as gene expression data at the mRNA level. However, these data do not provide direct information on the levels of proteins or their states of modification. Due to the recent advancement in the field of biological MS in conjunction with protein/DNA-sequence database search algorithms for the identification of proteins from MS data, it is possible to identify proteins with unprecedented speed and accuracy. However, it still remains a challenge to obtain quantitative information regarding the levels of identified proteins as well as site-specific modifications and their extent within individual protein molecules. In the sections below, we will describe different methods used in proteomics for the quantitative comparison of individual proteins present in cell pools that differ in some respect from one another (such as normal vs. disease) and for accurately determining changes in the levels of posttranslational modifications (such as phosphorylation, glycosylation) at specific sites of the individual proteins. Some of these methods can be applied directly to mixture of proteins and do not require a complete separation of individual protein components.
Proteomics and Systems Biology: Application in Drug Discovery and Development
7
The widely used methodologies for proteomics technology in general depends on (a) protein and/or peptide separation using two-dimensional-gel electrophoresis (2D-GE) or liquid chromatography, and (b) protein identification using MS. All these procedures are highly dependent on computer technology such as software for the image analysis of 2D gels, automated spot picking from the 2D gels, DNA or protein sequence database searches, proteome databases, and protein interaction maps (8). A third method using protein microarrays is also becoming a widely used proteomic technology for quantitative proteomics. In addition, surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS)-based ProteinChip® System has also been used for proteomic studies. 3.1. Two-Dimensional Gel Electrophoresis and Mass Spectrometry 3.1.1. Two-Dimensional Gel Electrophoresis
2D-GE has become a key technology in proteomics because of its high resolution as well as the ability to detect proteins with different kinds of posttranslational modification. The technique of 2D-GE was first introduced by Margolis and Kenrick (9) and later modified by O’Farrell (10) and Klose (11). In 2D-GE, proteins are separated by two orthogonal parameters, isoelectric point (pI) and molecular weight. The initial step of separation, isoelectric focusing (IEF) separates proteins based on differences in pI (the pH at which the net charge of any protein molecule is zero). This step uses precast immobilized pH gradient (IPG) strips and is available from vendors such as GE-Health Care Life Sciences and Bio-Rad Laboratories. The second dimension is sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE), which separates proteins based on their molecular weight; precast polyacrylamide gels with a variety of chemistries and gradients are commercially available from different vendors such as Bio-Rad Laboratories. There are some limitations of 2D-GE, such as poor reproducibility due to gel-to-gel variation, difficulty in automation, as well as poor ability to detect the low abundant proteins, hydrophobic proteins and proteins with very high or very low molecular weight and pI. The problem encountered in detecting low abundant proteins can be overcome by choosing an appropriate technique that enriches low abundant proteins prior to separation. The method of 2D-GE is widely used to separate proteins from two (or more) different samples followed by the visualization of separated protein spots with an appropriate stain. Following staining, the 2D gel images are captured using a charge couple device-based system or a laser scanner. The identification of protein spots which have undergone differential expression or posttranslational modification is carried out using computer-assisted appropriate image analysis software (such as Image Master Platinum 2D, GE Healthcare Life Sciences; PD Quest, Bio-Rad Laboratories; Progenesis SameSpots, Nonlinear dynamics, etc). For expression proteomics, different fluorescent stains such as SYPRO
8
Chakravarti, Mallik, and Chakravarti
Ruby (Bio-Rad Laboratories), Deep Purple (GE Healthcare Life Sciences), Flamingo (Bio-Rad Laboratories); as well as silver stain (Bio-Rad laboratories); and Biosafe coomassie (Bio-Rad Laboratories) are used. Fluorescent stains are also available for the detection of posttranslational modification present in glycoproteins (Pro-Q Emerald 300 glycoprotein stain, Molecular Probes) as well as phosphoproteins (Pro-Q Diamond Phospho protein gel stain, Molecular Probes). The stains used for 2D-GE should be compatible with the mass spectrometer since the proteins of interest are typically identified by MS following their separation by 2D-GE. Although a number of improvements have resulted in an increased reproducibility of proteome pattern between different laboratories using the 2D-GE technique, some major concerns still remain, for example, about the inability to resolve all the proteins of interest present in a biological sample. Two-dimensional difference gel electrophoresis (2D-DIGE) (12) is a relatively new technique in 2D-GE for a multiplex quantitative analysis of the component proteins of related but different protein samples (see review by Chakravarti et al. (13)). Commercially available CyDye DIGE Fluor minimal dyes (CyDye DIGE Fluor Cy2 minimal dye, CyDye DIGE Fluor Cy3 minimal dye and CyDye DIGE Fluor Cy5 minimal dye), with similar structures but different spectral characteristics, are widely used for 2D-DIGE. All of the CyDye DIGE fluor minimal dyes contain N-hydroxy succinimidyl ester group which forms covalent bonds with the e-amino groups of the lysine residues of a protein through an amide linkage. In the DIGE technology, prior to electrophoretic separation, different protein samples are labeled with different fluorescent dyes, pooled together, and the proteins present in the pooled samples are separated by IEF followed by SDSPAGE. Thus, it is possible to coseparate and codetect different but usually related protein samples on the same gel (12, 14). The technique of 2D-DIGE is carried out in four different steps: Step (1) – usually, two different but related sets of protein samples (such as normal and diseased tissue extract) are labeled with Cy3 and Cy5 separately. A pooled protein sample consisting of the same amount of all individual samples for one particular set of experiment is labeled separately with Cy2 and this is used as the pooled internal standard. Step (2) – separation using 2D-GE of the mixture of one set of samples, that is, one protein sample labeled with Cy3, one protein sample labeled with Cy5, and one pooled internal standard labeled with Cy2 is carried out on the same gel. Step (3) – the same gel is scanned three times with fluorophore-specific (Cy2, Cy3, and Cy5) excitation and emission wavelength and images are captured separately. Step (4) – image analysis and intragel quantitative comparison of all spots are carried out using the DeCyder Differential Analysis Software or Image Master Platinum 2D (GE-Health Care) or any other
Proteomics and Systems Biology: Application in Drug Discovery and Development
9
appropriate software such as PDQuest (Bio-Rad Laboratories) and Progenesis SameSpots (Nonlinear Dynamics). However, the DeCyder software is most widely used for DIGE image analysis. The use of the internal standard containing all the test samples ensures that all proteins present in the samples are represented, thereby allowing intragel (within the same gel) as well as intergel (within a number of gels) matching. Since the pooled internal standard is labeled with Cy2, intragel differences in protein abundances are determined by calculating the average ratio (Cy3:Cy2):(Cy5:Cy2). Inter gel comparison is carried out to find out the statistical significance of variation of any particular protein spot between different samples. Contrary to the traditional 2D-GE, the application of 2D-DIGE is restricted to protein expression and not for posttranslational modification of the proteins. For DIGE analysis, in addition to CyDye DIGE Fluor minimal dyes used for minimal labeling, CyDye DIGE Fluor saturation dyes are available which consist of two different dyes namely CyDye DIGE Fluor saturation dye Cy3 and CyDye DIGE Fluor saturation dye Cy5. These dyes have no intrinsic charge, contain thiol reactive maleimide groups and label all cysteine sulfhydryl groups in a protein sample. It is particularly useful for labeling protein samples available in limited quantity. Saturation labeling is much more sensitive than minimal labeling, as more fluorophore is incorporated into each protein species. However, proteins that do not contain cysteine cannot be labeled using saturation dyes. Since in saturation labeling, a Cy2 fluor is not available, the internal standard is labeled with one of the CyDyes and samples are labeled with the other CyDye. As a result of this, for similar experiments, saturation labeling requires twice as many gels as minimal labeling. For saturation labeling of a new sample, it is important that labeling conditions must be optimized to ensure the complete reduction of disulfide bonds and stoichiometric labeling of cysteine residues. 3.1.2. Mass Spectrometry
Several review articles on MS-based proteomics are available (15–19). Protein identification by MS can be carried out by the analysis of the whole protein (“top-down” proteomics) or that of peptides obtained from enzymatic or chemical cleavage of the protein (“bottom-up” proteomics). For either the “top-down” or the “bottom-up” approach, it is crucial to separate the proteins and/or peptides, as applicable, using different methods of separation, such as reverse phase, ion-exchange, size-exclusion, IEF, etc. Many of these techniques can be carried out either off-line or online with the mass spectrometer. Mass spectrometer measures the mass/charge (m/z) ratio of the gas phase ions. The mass spectrometer consists of an ion source, a mass analyzer and a detector. The ion source converts
10
Chakravarti, Mallik, and Chakravarti
the analyte molecules into gas-phase ions, the mass analyzer separates the analyte ions according to their m/z ratios, and the detector detects and records the number of ions at each m/z value. In proteomics research, proteins and/or peptides are ionized by any one of two different soft ionization techniques: (a) electrospray ionization (ESI) (20), and (b) matrix-assisted laser desorption/ionization (MALDI) (21). There are four different types of mass analyzers which are commonly used for proteomics research. They vary in their physical principles and analytical performance – quadrupole ion trap (QIT), linear ion trap (LIT) or linear trap quadrupole (LTQ), time of flight (TOF), and Fourier transform ion cyclotron resonance (FTICR). There are several hybrid instruments which combine the capabilities of different mass analyzers such as Q (quadrupole)-TOF, TOF–TOF, and LTQ-FTICR. Orbitrap is a new type of Fourier transform mass analyzer with high resolution, mass accuracy and dynamic range which is available in hybrid instruments such as LTQ Orbitrap (Thermo Fisher Scientific). Tandem mass spectrometry (MS/MS) is a widely used technique for protein identification in which mass spectra of peptide fragment ions are obtained to determine amino acid sequence information and posttranslational modification. Among the different fragmentation techniques, collision-induced dissociation (CID) is used most widely. More recent fragmentation techniques such as electron capture dissociation (ECD) (22, 23) and electron transfer dissociation (ETD) are used as well (24, 25). A widely used application of MS in quantitative proteomics is the identification of differentially expressed proteins which uses stable isotope tags to distinctly label proteins from two different conditions. In this method, the proteins present in two sets of similar samples are first separately labeled with different isotopes, combined together and then digested to yield labeled peptides. The peptides containing two different labels from the two sets of samples are separated by multidimensional liquid chromatography followed by their analysis using MS/MS. Proteins are identified by automated database searches of corresponding peptide MS/MS data and relative protein abundances are obtained from the intensity of the ion current in the mass spectra. Isotope coded affinity tag (ICAT) reagents (26) are the most commonly used isotope tags for this purpose and consist of three main components (a) a reactive group (iodoacetic acid) with specificity for cysteine residues; (b) a linker labeled with either light hydrogen (d0) or heavy hydrogen (such as eight deuterium: d8) isotope; and (c) an affinity tag (biotin) for the solid-phase capture and isolation of labeled peptides. Two different samples containing similar protein mixtures obtained from two different conditions are labeled separately with either the hydrogen-containing reagent or the deuterium-containing reagent. The two samples are then
Proteomics and Systems Biology: Application in Drug Discovery and Development
11
mixed together, digested with trypsin, and isotopically labeled peptides are purified by avidin affinity chromatography. The purification step eliminates noncysteine-containing peptides and thus reduces the complexity of the sample to be analyzed. The dis advantage of this method is that cysteine residues might be absent in the proteins of interest or might not be available for alkylation using cysteine-specific ICAT reagent due to posttranslational modifications of these residues. However, by using alternative chemistries, other amino acid residues can also be labeled with a stable isotope and analyzed (27–29). While this leads to different coverage if used alone, additional coverage can be obtained if the data obtained are combined with that from the ICAT approach. For example, Goodlett et al. (27) used permethyl esterification of peptides for the relative quantification of two different protein extracts and automated de novo sequence derivation of the same dataset. The method did not require the presence of cysteinecontaining proteins in the extract. Each of the protein mixture was digested with trypsin separately and then methylated with d0- or d3-methanol. By this process, carboxylic acid residues present on the side chain of glutamic acid, aspartic acid, and the carboxyl terminus were converted to their corresponding methyl esters. The separate mixture of d0- and d3-containing peptides were combined and analyzed by liquid chromatography tandem mass spectrometry (LC-MS/MS). Following correlative database searching of peptide tandem mass spectra, the parent proteins of the methylated peptides were identified. It is possible to obtain ratios of the proteins in the two original mixtures by normalization of the area under the ion current intensity curve for identical charge state of d0- to d3-methylated peptides. These investigators also developed an algorithm to derive peptide sequence de novo by the comparison of tandem mass spectra of the d0- and d3-peptide methyl esters. In another approach (28), using 2-methoxy-4,5-dihydro-1H-imidazole, a lysine specific labeling reagent, it was possible to convert the lysine residues at the C-terminus as well as tryptic digests of the proteins to 4,5-dihydro1H-imidazole-2-yl derivatives. Since the mass spectra of the derivatized peptides show greater number of more intense features than their underivatized counter parts, more information could be obtained from peptide-mapping experiments. Also, this label can be produced as stable isotopic forms containing four deuterium atoms which can be used for differential quantitative studies. Several improvements made in isotope tagging and data analysis have made stable isotope tagging a useful technique for quantitative proteomics studies. Interestingly, some other chemical reactions have also been used to introduce tags into specific sites of peptides or proteins to probe specific functions of proteins. For example, phosphorylated peptides have been isolated using isotope labeling and selective
12
Chakravarti, Mallik, and Chakravarti
chemistries to enrich phosphoproteins from a complex mixture (30–34). Since level of phosphorylation plays an important role in several disease conditions, the quantitative comparison of this posttranslational modification in control and disease cells is helpful to monitor disease progression, cure as well as the identification of putative drug candidates. In fact, phosphoproteomics has been used as a new research front leading to drug discovery. It is estimated that approximately one-third of the total cellular proteins undergoes posttranslational reversible phosphorylation and play important roles in the regulation of cellular signaling network in response to external stimuli and signal transduction mechanisms (35–37). Dysregulation in the processing of these signals has been implicated to be associated with human pathologies, including many forms of cancers (36, 37). Thus, it is of no surprise that a new research front is being developed for highthroughput structure-function profiling of phosphoproteomes, and the information is used to find effective drug targets as well as for the development of early detection of disease biomarkers. In eukaryotes, serine residues are the most common sites of phosphorylation (~90%), followed by threonine (~10%) whereas tyrosine residues are rarely phosphorylated (~<0.05%) (37). Phosphoproteomics involves a high-throughput identification and quantitative analysis of the phosphorylation states of proteins in a given biological sample and relating them to the signaling events as well as understanding of the underlying biological implications. A database has been developed to maintain the information of all possible phosphorylated sites in all possible proteins (http://www.phosphosite.org). It already lists >30,000 phosphorylation sites on >17,000 proteins. As the analysis of new samples continues, the database will grow larger with time. We will briefly discuss the sample preparation prior to MS based high-throughput profiling of phosphoproteomes. Sample enrichment prior to MS is necessary particularly for the analysis of phosphoproteins. The dynamic range of proteomes and phosphoproteomes are in the same order (~109) (37), but the analysis of phosphoproteomes is difficult due to the substoichiometric and transient phosphorylation of these proteins. The goal of studying phosphoproteomes is to identify sites of phosphorylation, the quantification of such site stoichiometry, and monitoring the link between cellular perturbations and the temporal changes in phosphorylation levels. Many proteins that are involved in signaling networks are only sparingly expressed. Together with this, the transient phosphorylation/ dephosphorylation dynamics generate a pool of extremely low levels of phosphopeptides after proteolytic digestion in comparison to large number of total peptides in cell or tissue extract. Consequently, the MS signal intensities of phosphopeptides of interest are greatly suppressed and thus the efficacy of such
Proteomics and Systems Biology: Application in Drug Discovery and Development
13
method is questionable. Therefore, the enrichment of the phosphorylated proteins and peptides in the sample before MS is necessary. Currently a number of phosphopeptide enrichment strategies are followed based on the principle of immunoprecipitation and some specialized chromatographic techniques. A number of recent review articles are available that discuss in detail about these methods including the choices of one method over the other depending on the final objective (33, 37, 38). For example, immunoprecipitation using antiphosphotyrosine antibodies is generally applied to enrich phosphotyrosine proteins and peptides since compared to phosphorylated serine and threonine residues, the occurrence of tyrosine phosphorylation is very low. On the other hand, antibodies specific to phosphoserine and phosphothreonine are not very effective for their enrichment and therefore are not used routinely. For global qualitative phosphoproteome profiling, the most common technique used for the enrichment process is the immobilized metal affinity chromatography (IMAC). It depends on the high affinity of phosphate groups for metal ions such as Fe3+, Zn2+, and Ga3+. Although the highly negative environment around phosphate groups have higher preference to bind with strong metal cations, weakly acidic peptides may also bind with the metal ions leading to the enrichment of nonphosphorylated acidic peptides. Consequently, during IMAC, nonphosphoproteins and peptides will be eluted along with phosphorylated substrates as well. However, a multistep separation and/or mixture of multiple metal ions could be employed to reduce the heterogeneity of the enriched proteins and peptides. Also, the conversion of carboxylate groups to ester groups can eliminate this nonspecific interaction (34). The titanium dioxide-based metal oxide affinity chromatography (MOAC) is another form of IMAC with some added advantages, such as shorter preparation time and increased capacity compared to the IMAC resin. However, nonspecific retention of nonphosphorylated peptides is also a problem for this technique. Other methods such as strong cation exchange (SCX), a blend of anion and cation exchangers (ACE), are also used for the selective enrichment of phosphopeptides. In practice, generally multistep enrichment strategy is followed using a combination of the above techniques. The development of technically less complicated and less expensive enrichment strategies is expected in future. The quantification of phosphoproteins has important application in clinical research. Changes in phosphorylation level before and after a cellular perturbation are linked with the specific biological response. Generally, phosphorylated samples are labeled with nonradioactive stable isotopes through chemical modifications or through the addition of labeled amino acids in cell culture. Label-free methods are also used. However, recent developments in this field have mostly been devoted to the effective enrichment
14
Chakravarti, Mallik, and Chakravarti
strategies and qualitative identification of global or specific phosphorylation sites rather than the quantitation aspects. One of the most used quantitation methods in the multiple chemical modification to incorporate stable isotopes is “isobaric tag for relative and absolute quantitation” (iTRAQ). The isobaric tags are a group of designed chemical reagents, each one consisting of a “reporter region,” a “balance region,” and an amine-specific “peptide reactive group region” (iTRAQ™ Reagents, Applied Biosystem, http://www.appliedbiosystems.com). In a specially designed and controlled MS/MS fragmentation method, the bonds between the reporter region and the balance region as well as that between the balance region and the reactive group region are broken along with the usual peptide fragmentation. In a fourplex version (currently eight-plex is also available) of the iTRAQ reagents, the individual masses of the reporter region and the balance region are different, but the combined mass is maintained constant. Therefore, even if a digested peptide fragment is tagged with four such reagents, the MS signal will not split. However, during MS/MS, one can locate the distinct reporter signals which are designed in a way such that these signals do not interfere with peptide fragment ions. The peak area determined for the reporter ions can quantify the corresponding peptides originating from different samples. In addition, the MS/MS ion pattern or signature will allow the identification of the protein from sequence database using suitable algorithm. Thus, iTRAQ is a method for the simultaneous identification and quantitation of proteome including the phosphoproteome. Another powerful quantitative technique is “stable isotope labeling with amino acids in cell culture” (SILAC). A specific cell type is grown in two different media, one containing natural amino acids and the other containing an isotopic form. All proteins expressed in two different media will be expected to incorporate either natural or isotope-containing amino acids. Following enzymatic digestion and mixing of the samples, each peptide is expected to generate a pair of two peaks with difference in mass based on the isotope type in MS run. The different peak intensities for such pairs can give an estimate of the relative quantities of a particular protein compared to others. Significant isotope incorporation requires multiple cell divisions and because of this reason SILAC is not a suitable method of quantitative proteome analysis for the comparison of primary cells or tumor tissues. Different advantages and disadvantages and precautionary measures required at different stages have been discussed in detail by Nita-Lazar et al. (37). For absolute quantitative estimation, it is necessary to add an internal peptide to the sample (39). 3.2. Protein Arrays
DNA microarrays enable the measurement of mRNA expression level of thousands of genes in a single experiment. Similarly,
Proteomics and Systems Biology: Application in Drug Discovery and Development
15
protein arrays can be used to achieve high throughput protein analysis in parallel and is becoming a key proteomics technology. Protein arrays are solid-phase ligand binding assay systems using immobilized protein spots on surfaces such as glass, membranes, microtiter wells, etc. Bioinformatics support is important for data handling. The protein spots can be homogenous or heterogeneous and may be antibody, a cell or phage lysate, nucleic acid, drug, recombinant protein or peptide, etc. Protein arrays can provide valuable information to detect proteins, monitor their expression levels, posttranslational modification, and functions as well as investigate protein–protein interactions. Using protein arrays, it is possible to carry out parallel multiplex screening of multiple interactions, such as protein–antibody, protein–protein, protein–ligand or protein–drug, enzyme–substrate screening, etc. Usually a very small amount of sample is required. Although used widely, 2D-GE and MS-based proteomics techniques may not be able to detect low abundant proteins. This demands other sensitive and easily accessible high throughput technologies for the detection and differential expression of proteins in health and disease. Protein array technology is useful for such purposes and can be applied to protein expression profiling ranging from limited numbers of proteins to global proteomic analysis. It can be used in target identification as well as in the validation process. In general, 2D-GE and MS-based techniques can be used for the initial identification of biomarkers and subsequently specific antibodies can be used in protein microarrays for rapid high throughput screening of a large number of samples. These techniques complement each other. Protein microarray formats can be divided into two major classes – forward phase arrays and reverse phase arrays (RPA). In the forward phase array format, the analyte(s) of interest is present in solution phase and the capture molecules (usually antibodies) are immobilized on the solid surface. Usually each array measures one test sample such as a cellular lysate or serum sample representing a control or particular disease condition and multiple analytes are measured simultaneously. On the contrary, in the RPA format, the array may contain samples from a number of patients or from the same patient at different stages of disease or treatment and each array is incubated with one detection protein (e.g., antibody). The single analyte concentration can be compared among different patients or from same patients with different state of disease progression or different time points of drug treatment. The detection of the array is carried out using a tagged antibody, ligand or cell lysate or serum. The signal intensity is proportional to the binding of the tagged molecule bound to the analyte molecule (40). The multiplex analysis of the same spot for the detection of multiple analytes is possible using dual color infrared dye-labeled antibodies as well as quantum dots (41, 42).
16
Chakravarti, Mallik, and Chakravarti
For yeast (Saccharomyces cerevisiae), the whole proteome array which consisted of 5,000 purified proteins were deposited onto a glass microscope slide. This was used for global studies on protein–protein and protein–lipid interactions (43). In spite of such success, it is important to remember that protein chips arrays are much more challenging than DNA arrays due to the following reasons – the solubility of the protein can vary widely, their expression levels have a broad dynamic range, and proteins are less stable than DNA. In addition, it is generally difficult to preserve the native structures of proteins on the glass slides. In contradistinction, for DNA microarrays, only the nucleotide sequence needs to be maintained. While used in combination with laser capture microdissection (LCM) (44) technology, RPA can be particularly useful for gaining information, particularly about cell signaling molecules and posttranslational modification from tissue sections obtained from normal and diseased individuals. In fact, the RPA platform has been used to explore a variety of signaling pathways involved in malignant progression and tumor biology (4). For example, preliminary data indicate that protein kinase C is downregulated during the progression of prostate cancer in a study of microdissected cells from normal, stroma, and prostate tumors (45). Prostate tumors, like most tumors, are solid tumors that grow as solid masses of tissue and consist of two distinct but interdependent compartments: the parenchyma containing neoplastic cells and the stroma that the neoplastic cells induce and in which they are dispersed. If validated, this finding on protein kinase C and prostate cancer could have profound effects on the rationale behind some current therapies (46). Also, by applying RPA in normal prostate epithelium, prostate intraepithelial neoplasia, and invasive prostate cancer tissues (45, 47), it has been possible to identify specific molecular changes like changes in signaling molecules, such as Akt, GSK3B, PKC-R, and p38, that are implicated with cancer progression between different tissues. For these studies, the cancerous tissues were isolated by LCM, arrayed on nitrocellulose membranes, and probed with appropriate antibodies. These examples illustrate the importance of proteomics technology coupled to signal pathway profiling in providing novel insights into the cellular processes. In fact, this technique can be applied to study normal physiology and pathology of any tissue at the level of the proteome by comparing diseased and healthy tissues within the same patient which will be helpful for the development of individualized diagnosis and treatment strategies. Extreme sensitivity for the detection of an analyte is an advantage of this technology – in fact, attomole level of a protein can be detected. Using this technology, it is possible to analyze signaling pathways using small numbers of cultured cells or cells isolated by
Proteomics and Systems Biology: Application in Drug Discovery and Development
17
LCM from human tissue which were available during clinical trials (45, 47–49). As pointed out by Sheehan et al. (50), different probes, such as antibodies, aptamers (oligonucleotides or peptides that bind to specific molecular targets such as proteins), ligands and drugs, required for protein microarrays, often cannot be obtained with predictable affinity or specificity. In many instances, it is difficult to obtain highly specific antibodies or suitable protein-binding ligands. Since the turning on or off of signaling pathways depend on posttranslational modifications as well as protein–protein interactions, it is important to have specific antibodies or probes that are specific for the detection of such modifications or interactions. It is strongly recommended that the specificity of the antibodies should be thoroughly assessed and validated by Western blot analysis before using those antibodies in protein array format. Such validation may include the demonstration of single band of appropriate molecular weight in Western blot analysis of a complex biological mixture similar to the one to be analyzed by microarray format. For a phosphospecific antibody, it may be validated by differential detection of protein bands from lysates prepared from control and activated cells in which the signaling pathway is known to be activated. It will be helpful to have a realtime validation of the antibodies by printing out negative and positive controls in each array. A web posting of a set of validated antibodies can be found at the www.home.ccr.cancer.gov/ ncifdaproteomics/. Cooperation among different funding agencies and international consortia will aid in the generation of large comprehensive libraries of fully characterized specific antibodies, ligands, and probes. Interestingly, such initiatives are already in progress by the Human Proteome Organization (www.hupo.org) and the Human Proteome Resource in Sweden (www.proteinatlas.org/intro.php). Individual investigators are also beginning efforts to provide the scientific community with critical antibody resources (51–53). A reagent resource to identify proteins and peptides of interest for the cancer community has been the topic of a workshop arranged by NCI (54). We would like to point out that the global ICAT technology has certain advantages over protein chip array when it is necessary to quantify a large number of proteins. As one can understand, the generation of antibodies, proteins or protein capture agents as well as their deposition on the chip can be time-consuming, expensive, and technically challenging. Also, the capacity for accurate quantification can be problematic. 3.3. SELDI-TOF-MSBased ProteinChip® System
Protein quantification can also be carried out using ProteinChip® array which depends on selective binding of proteins according to their charge or hydrophobicity from a complex biological mixture such as plasma followed by accurate and sensitive detection by
18
Chakravarti, Mallik, and Chakravarti
SELDI-TOF-MS (55). Bio-Rad Laboratories, Hercules, CA (www.bio-rad.com) is currently the source for SELDI technology products for the life science market, and Vermillion, Inc.(www. vermillion.com) is the source of these products for the molecular diagnostics market (www.vermillion.com/alliances_and_partnerships.cfm). SELDI allows the capture of proteins of interest on a chip and their subsequent analysis essentially by MALDI-TOF MS. This technology enables the comparative analysis of virtually any given protein-containing solution in a fast and simple process and requires only a minute amount of samples. A complex mixture of proteins can be applied to the spots of ProteinChip Arrays, which have been derivatized with specific chromatographic chemistries. The proteins interact and bind with the chromatographic array surface according to their interaction and binding potential. Subsequent on-spot washing removes salts or other unbound molecules. This on-chip retentate chromatography step thus leads to fewer background peaks that might interfere with the detection of low abundance species during MALDI-TOF MS. In fact, protein interaction studies or enzymatic reactions may be carried out directly on-spot under physiological conditions. The chromatographic surfaces provide a convenient support for the cocrystallization of matrix and target proteins, resulting in the formation of a homogenous layer on the spot which makes them ideal for subsequent MALDI-TOF analysis. It has various applications, including biomarker discovery and assay development (56), protein interaction studies (57), monitoring of enzymatic reactions (58) and for process proteomics approaches (59). Different kinds of quantification experiments have demonstrated the high sensitivity and reproducibility of the SELDI System. For example, the quantitative detection of myoglobin from 1% blood serum, the simultaneous capture and detection of IL-8 and epidermal growth factor from 20% blood serum, and the quantitative monitoring of phosphatase activity over time.
4. Applications of Phosphoproteomics in Drug Development
Quantitative phosphoproteomics provides useful tools for the development of target-oriented drug discovery and early diagnostic biomarkers. Multiple biochemical pathways leading to cancer initiation, development, and progression are in general due to irregular cellular signaling and thus involve an altered role of protein kinases (38). The identification and drug-induced alteration of the defects in any one or multiple such pathways may slow down the growth of cancer cells. Moreover, specific kinases are assumed to be the signatures for certain types of cancers.
Proteomics and Systems Biology: Application in Drug Discovery and Development
19
This would help disease diagnosis as well as grouping patients that might require personalized therapy. Therefore, in recent years, much attention has been paid to the development of kinasespecific drug targets to treat cancers. Quantitative phosphoproteomics is considered to be an appropriate and potential method to study kinases and their roles in cancer therapy and diagnosis and a number of kinase-based approved anticancer therapeutics have been discussed by Yu et al. (38). The phosphorylation status of multiple proteins can also be effectively used to map the response of host cells to infection and thus can be useful to develop drugs for infectious diseases.
5. Toxicoproteomics and Its Application in Drug Development and Clinical Management
5.1. Hepatotoxicity
In the postgenomic era, proteomic techniques have been used to study the changes in protein profile following exposure to environmental toxins and drugs. This area of research is known as toxicoproteomics. The application of proteomics in toxicology can be divided into two broad overlapping areas: (a) mechanism of toxicity, and (b) screening and predictive toxicology. If it is possible to establish a relationship between toxic effects and protein markers and develop a database of such markers, it will be useful to screen a new drug for its potential toxic effects against a panel of predictive toxicity markers and will be an early indication of toxicity for further investigation. In the past, toxicity associated with any experimental drug was identified in animal studies according to the recommendation of the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH; www.ich.org) and was dependent mostly on histopathological and biochemical parameters. The early detection of potential toxicity of potential drug candidates may cause significant reduction in time and cost for a new drug to become available in the market. Liver is the major site of metabolism and detoxification in the body and often the target of drug-induced toxicities. Consequently, hepatotoxicity is used to predict adverse drug reactions. Druginduced liver toxicity can be metabolism based (predictable) or idiosyncratic (unpredictable). Traditionally histopathological parameters such as hepatocellular vacuolation (steatosis) and periductual inflammation were used to measure the liver toxicity potential of investigational drugs in preclinical studies (60). Additionally, biochemical parameters such as elevated plasma level of secreted liver enzymes, such as alanine aminotransferase and aspartate aminotransferase were considered to be indicative of
20
Chakravarti, Mallik, and Chakravarti
liver injury (61). But these methods cannot be applied for in vitro studies and fail to predict idiosyncratic toxicities. Recently, the use of proteomic techniques in predicting hepatotoxicity has been used by different laboratories. Examples include the development of rodent liver proteomic toxicity database based on studies on the effects of a range of xenobiotics on protein expression in the liver (62). Methylpyrilene, a widely used antihistamine was discontinued due to its carcinogenic property and was replaced by structurally similar pyrilamine which also has the antihistamine property but is not a carcinogen. Interestingly, using 2D-GE, while the former compound was found to change liver proteome profile, pyrilamine did not lead to such change (63). Such observation points toward the potential role of proteomics in the selection of lead candidates. Studies have also been carried out on the hypoglycaemic compound SDZPGU693 which stimulates glucose utilization in peripheral tissues (64). Using 2D-GE, liver samples of treated and untreated rats identified change in expression levels of several proteins which gave insights into the molecular mechanisms of both its pharmacological action and a toxic response. Similar studies by other investigators (65, 66) were able to identify proteins which are part of pathways known to be disrupted in drug-induced liver steatosis and hence their possible role in hepatotoxicity. The downregulation of some of the secretory proteins suggested the disruption of secretory pathway as a mechanism of liver toxicity. Interestingly, such downregulation could be detected as early as 6 h of treatment with the experimental compound suggesting a window of time that occurs before the manifestation of clinical symptoms and that time period can be analyzed by proteomic technology to predict the hepatotoxicity. Using two-dimensional biphasic liquid chromatography and tandem mass spectrometry (LC/LC/MS/MS), immunochemical techniques, and immortalized normal human hepatocytes, Gao et al. (67) tested hepatotoxicity of 20 different drugs. They observed increased levels of secretory proteins BMS-PTX-265 and BMS-PTX-837. For all 20 drugs, the elevations of BMS-PTX-265 correlated exactly with the known safety profile; whereas changes in BMS-PTX-837 correctly predicted the safety profile in 19 of 20 drugs (one false negative). In summary, the data support the preclinical in vitro method as a means to identify new biomarkers of liver toxicity, as well as the validity of the biomarkers themselves. 5.2. Nephrotoxicity
Apart from hepatotoxicity, nephrotoxicity is also used in the evaluation of drug toxicity using proteomic technology. Renal tissue is a sensitive organ for drug-induced toxicity for various reasons such as large blood flow and the presence of a variety of xenobiotics transporters and metabolizing enzymes. In addition, drugs and metabolites become concentrated in the tubules during urine production. Traditional evaluation of nephrotoxicity includes the
Proteomics and Systems Biology: Application in Drug Discovery and Development
21
measurement of serum metabolites such as blood urea nitrogen and serum creatinine, (68) as well as the identification of specific urine markers such as gamma glutamyl transferase or N-acetyl glucosamine (69) but are of inadequate sensitivity. Bandara et al. (70) identified a correlation between proteomic evaluation and conventional measurements in the assessment of renal proximal tubular toxicity. Using current proteomics technology of 2D-GE and MS, they identified the level of T-kininogen (a cysteine protease inhibitor expressed in kidney) to be elevated in rats following the treatment with cisplatin (and other known nephrotoxins) at early time points returning to basal level after 3 days of treatment suggesting T-kininogen may be required to counteract apoptosis in proximal tubular cells in order to minimize tissue damage following a toxic insult. Cycosporin A is an immunosuppressive drug whose use is restricted by adverse effects. Following the proteomic analysis of kidney homogenate of CsA-treated rats, vitamin D-dependent calcium-binding protein calbindin-D (28K) was identified as a novel biomarker for CsA-induced renal toxicity (71). Interestingly, while rats and humans displayed CsA-induced renal cytotoxicity, monkeys and dogs did not. In accordance with this phenomenon, calbindin-D was also regulated in a speciesspecific manner (72). Gupta et al. (73) postulated a toxicoproteomics-based new drug development paradigm following appropriate target identification. It was suggested that the information obtained on drug toxicities from preclinical studies should be used throughout the development process as well as after regulatory approval as part of postmarketing surveillance. Also, at each stage of the process, the requirement for multiplexing decreases while that for specificity and sensitivity increases. In essence, high-throughput proteomics tools are necessary in the preclinical and early clinical trial stages for screening the proteome for biomarkers. However, once the relevant markers have been identified, then accurate and precise tests are necessary for clinical use. The suggested importance of toxicoproteomics at each stage of drug development is schematically represented in Fig. 1 and is described below: 5.3. Lead Optimization
This involves the chemical modification of a biologically active compound to fulfill all stereoelectronic, physicochemical, pharmacokinetic, and toxicologic properties required for clinical usefulness. In vitro assays using different cell lines, proteomics techniques such as 2D-GE-MS/MS or LC-MS/MS can be applied to find out adverse cellular changes associated with each drug candidate.
5.4. Preclinical Studies
Such studies are carried out in animal models to find out the in vivo safety, efficacy, and maximal tolerated dose. Using proteomics techniques such as 2D-GE-MS/MS or LC-MS/MS,
22
Chakravarti, Mallik, and Chakravarti
Stage of Drug Development
Screen out compounds with unacceptable cellular phenotype changes.
Preclinical Toxicology Studies
Predict potential human toxicities. Identify biomarkers of toxicity. Identify Maximum Tolerated Dose (MTD).
2D-GEMS/MS or LC-MS/MS
2D-GEMS/MS or LC-MS/MS
Identify individual variations in biomarkers of toxicity.
Phase I Clinical Trials
Optimize dosing schedule.
Protein Arrays SELDI-TOF
Develop tests to identify potential ‘idiosyncratic’ reactions.
Phase II/III Clinical Trials
Clinical Management
Time-series experiments to identify potential ‘cumulative’ toxicities – optimize maximum duration of therapy. Periodic toxicoproteomics monitoring. Identify optimum dose for each individual.
Protein Arrays SELDI-TOF
Decreasing need for multiplexing
In vitro Assays
Applicable Proteomics Techniques
Increasing need for sensitivity and specificity
Lead Optimization following Target Identification and Validation:
Application of Toxicoproteomics
Protein Arrays Immunohistochemistry ELISAs
Fig. 1. New drug development paradigm in the context of toxicoproteomi. Drug development is moving toward a personalized approach with the identification of biomarkers for screening patients likely to have dose-related or idiosyncratic toxicities from an investigational compound. In the new paradigm, proteomic tools will be used extensively in preclinical and clinical studies to identify these predictive biomarkers. Diagnostic tests based on these markers may then be used in clinical medicine to predict drug-related toxicities. Reprinted from ref. (73) with permission from Bentham Science Publisher Ltd.
it may be possible to find out biomarkers as well as the mechanism of tissue toxicity and predict the possible toxic effect of a potential drug candidate in humans. 5.5. Phase I Clinical Trials
Following successful preclinical studies, Phase I trial is carried out in human populations to find out the safety of a potential drug candidate in humans as well as to find out one or more appropriate dosing schedule. Using proteomics technology (protein arrays
Proteomics and Systems Biology: Application in Drug Discovery and Development
23
and SELDI-TOF), it may be possible to identify markers indicative of drug-related toxicity. Such markers can be identified in serum, urine, leukocytes, or easily available tissues such as oral mucosa (74, 75). This can be helpful to personalize individual dose based on the level of these marker molecules in different individuals. It can also be helpful to find out the idiosyncratic reaction observed in any individual. 5.6. Phase II/III Clinical Trials
Following a successful Phase I trial, a Phase II or Phase III clinical trial would be helpful for the identification of markers of longterm or cumulative toxicity. This, in turn, helps further optimization of the dose of the drug, dosing frequency as well as maximum duration of the therapy that may be applicable to the patient.
5.7. Clinical Management
Use of toxicoproteomics will be particularly helpful for personalized medicine. Before treatment, the patient can be monitored for various biomarkers for dose-related or idiosyncratic toxicities. This will help in the choice of therapy as well as dose and the frequency of the drug to be given to the patient. Even during treatment, periodic monitoring for the toxic biomarker will be helpful to change the type of therapy or the dose or frequency of the drug.
6. Conclusion Due to the tremendous rapid progress in proteomics and genomics technologies, it is possible to carry out detailed in-depth characterization of the molecular basis of different cellular functions contributing to normal or pathological cellular activities. In the past, drug discovery was mostly focused on a particular protein which could be connected to a particular disease. In recent years, there has been a shift toward strategies for the application of systems-oriented approaches. Such approaches could allow more effective therapeutic intervention based on a systems-oriented understanding of the disease biology and drug response. In essence, the emerging fields of systems biology and proteomics are offering new ways of diagnosis and prognosis of various disease as well as personalized medicine. Significant progress has been made in cancer research using systems biology and the proteomics platform. However, there still remains a lot of work to be done for the proper unraveling and understanding of systems biology. The understanding of protein and gene regulatory networks of biological systems will improve drug development efforts and prevention of diseases. As pointed out by Weston and Hood (1), targeting the key nodal points of networks will help circumvent the disease potentials emerging from defective genes
24
Chakravarti, Mallik, and Chakravarti
(somatic or inherited) or pathological environmental stimuli. These nodes may therefore be more effective targets for therapeutic interventions. Bioinformatics play a major role since a large volume of different types of data need to be collected and integrated for the proper understanding of biological systems for accurate diagnosis and prognosis during drug therapy. Information technology and computational support are an essential part of such efforts. While the expression and activities of several disease-associated molecules can provide important information, a full understanding of the participation and interaction of these molecules within the cellular control networks is equally important. It is important to understand the correlation of a drug target with complex cellular networks involved with the disease process in order to obtain high efficacy at low drug dosage and reducing the probability of development of drug-resistance. Inhibition of the activity of a signaling protein molecules without understanding its role in the overall cellular mechanism, can ultimately lead to an overall adverse effect. The recent advancement in the proteomics platform would help enormously to achieve this goal. It is worth mentioning that although resources and technologies are increasingly available to identify and evaluate the efficacy and safety of different chemical compounds, such improvement has not been reflected in the approval of drugs against new targets – that is cellular molecules not perturbed or targeted by previous drugs. In spite of significant increase in research and development expenditure, there is a disproportionately low number of new drug targets. One of the examples of major success in this area includes the small molecule ABL kinase inhibitor imatinib mesylate (Gleevec, Novartis) in the treatment of chronic myelogenous leukemia. Finally we would like to mention that ideal proteomics techniques for drug discovery require separation and characterization of the whole proteome of an organism under study. Other requirements include the identification of protein activity independent of protein abundance, and protein–protein and protein–small-molecule interactions. Much of the current proteomics research that is focused on target-protein identification would be impossible to carry out without the advancements in biological MS. Unfortunately, membrane proteins as well as low abundant proteins cannot be analyzed efficiently using some of the currently available separation and analytical techniques, such as 2D-GE. It is also necessary that these techniques should be user-friendly, automated, high-throughput, and cost-effective. Although enormous progress has been made in recent years, there still remains space for improvement in this area of proteomics research.
Proteomics and Systems Biology: Application in Drug Discovery and Development
25
Acknowledgments The authors would like to acknowledge the generous support provided by the Keck Graduate Institute of Applied Life Sciences, Claremont, CA, The Arnold and Mabel Beckman Foundation, the Ralph M. Parsons Foundation, and the National Science Foundation grant FIBR 0527023. References 1. Weston AD, Hood L (2004) Systems biology, proteomics, and the future of health care: toward predictive, preventative, and personalized medicine. J Proteome Res 3:179–196 2. Heller MJ (2002) DNA microarray technology: devices, systems, and applications. Annu Rev Biomed Eng 4:129–153 3. Griffin TJ, Gygi SP, Ideker T, Rist B, Eng J, Hood L, Aebersold R (2002) Complementary profiling of gene expression at the transcriptome and proteome levels in Saccharomyces cerevisiae. Mol Cell Proteomics 1:323–333 4. Szkanderová S, Port M, Stulík J, Hernychová L, Kasalová I, Van Beuningen D, Abend M (2003) Comparison of the abundance of 10 radiation-induced proteins with their differential gene expression in L929 cells. Int J Radiat Biol 79:623–633 5. Greenbaum D, Colangelo C, Williams K, Gerstein M (2003) Comparing protein abundance and mRNA expression levels on a genomic scale. Genome Biol 4:117 6. Anderson NL, Anderson NG (2002) The human plasma proteome: history, character, and diagnostic prospects. Mol Cell Proteomics 1:845–67, Erratum in: Mol Cell Proteomics. 2, 50 7. Wilkins MR, Pasquali C, Appel RD, Ou K, Jean-Charles Sanchez OG, Yan JX, Gooley AA, Hughes G, Humphery-Smith I, Williams KL, Hochstrasser DF (1996) From proteins to proteomes: large scale protein identification by two-dimensional electrophoresis and amino acid analysis. Nat Biotechnol 14:61–65 8. Chakravarti DN, Chakravarti B, Moutsatsos I (2002) Informatic tools for proteome profiling. BioTechniques Suppl:4–10, 12–5 9. Margolis J, Kenrick KG (1969) 2 dimensional resolution of plasma proteins by combination of polyacrylamide disc and gradient gel electrophoresis. Nature 221:1056–1057 10. O’Farrell PH (1975) High resolution two dimensional electrophoresis of proteins. J Biol Chem 250:4007–4021
11. Klose J (1975) Protein mapping by combined isoelectric focusing electrophoresis of mouse tissues. A novel approach to testing for induced point mutations in mammals. Humangenetik 26:231–243 12. Unlu M, Morgan ME, Minden JS (1997) Difference gel electrophoresis: a single gel method for detecting changes in protein extracts. Electrophoresis 18:2071–2077 13. Chakravarti DN, Gallagher S, Chakravarti B (2004) Difference gel electrophoresis: application in quantitative proteomics research. Curr Proteomics 1:261–271 14. Tonge R, Shaw J, Middleton B, Rowlinson R, Rayner S, Young J, Pognan F, Hawkins E, Currie I, Davison M (2001) Validation and development of fluorescence two-dimensional differential gel electrophoresis proteomics technology. Proteomics 1:117–124 15. Han X, Aslanian A, Yates JR (2008) Mass spectrometry for proteomics. Curr Opin Chem Biol 12:483–490 16. Guerrera IC, Kleiner O (2005) Analysis of mass spectrometry in proteomics. Biosci Rep 25:71–93 17. Domon B, Alving K, He T, Ryan TE, Patterson SD (2002) Enabling parallel protein analysis through mass spectrometry. Curr Opin Mol Ther 4:577–586 18. Yates JR (2004) Mass spectral analysis in proteomics. Annu Rev Biophys Biomol Struct 312:212–217 19. Domon B, Aebersold R (2006) Mass spectrometry and protein analysis. Science 31:212–217 20. Fenn JB, Mann M, Meng CK, Wong SF, Whitehouse CM (1989) Electrospray ionization for mass spectrometry of large biomolecules. Science 246:64–71 21. Karas M, Hillenkamp F (1988) Laser desorption ionization of proteins with molecular masses exceeding 10,000 daltons. Anal Chem 60:2299–2301
26
Chakravarti, Mallik, and Chakravarti
22. Zubarev RA, Kelleher NL, McLafferty FW (1998) Electron capture dissociation of multiply charged protein cations. A nonergodic process. J Am Chem Soc 120:3265–3266 23. Zubarev R (2006) Protein primary structure using orthogonal fragmentation techniques in Fourier transform mass spectrometry. Expert Rev Proteomics 3:251–261 24. Syka JEP, Coon JJ, Schroeder MJ, Shabanowitz J, Hunt DF (2004) Peptide and protein sequence analysis by electron transfer dissociation mass spectrometry. Proc Natl Acad Sci USA 101:9528–9533 25. Pitteri SJ, Chrisman PA, Hogan JM, McLuckey SA (2005) Electron transfer ion/ion reactions in a three-dimensional quadrupole ion trap: reactions of doubly and triply protonated peptides with SO2−. Anal Chem 77:1831–1839 26. Gygi SP, Rist B, Gerber SA, Turecek F, Gelb MH, Aebersold R (1999) Quantitative analysis of complex protein mixtures using isotope-coded affinity tags. Nat Biotechnol 17:994–999 27. Goodlett DR, Keller A, Watts JD, Newitt R, Yi EC, Purvine S, Eng JK, von Haller P, Peters EC, Horn DM, Tully DC, Brock A (2001) A novel multifunctional labeling reagent for enhanced protein characterization with mass spectrometry. Rapid Commun Mass Spectrom 15:2387–2392 28. Peters EC, Horn DM, Tully DC, Brock A (2001) A novel multifunctional labeling reagent for enhanced protein characterization with mass spectrometry. Rapid Commun Mass Spectrom 15:2387–2392 29. Munchbach M, Quadroni M, Miotto G, James P (2000) Quantitation and facilitated de novo sequencing of proteins by isotopic N-terminal labeling of peptides with a fragmentationdirecting moiety. Anal Chem 72:4047–4057 30. Oda Y, Huang K, Cross FR, Cowburn D, Chait BT (1999) Accurate quantitation of protein expression and site-specific phosphorylation. Proc Natl Acad Sci USA 96:6591–6596 31. Zhou H, Watts JD, Aebersold R (2001) A systematic approach to the analysis of protein phosphorylation. Nat Biotechnol 19:375–378 32. Goshe MB, Conrads TP, Panisko EA, Angell NH, Veenstra TD, Smith RD (2001) Phosphoprotein isotope-coded affinity tag approach for isolating and quantitating phosphopeptides in proteome-wide analyses. Anal Chem 73:2578–2586 33. Goshe MB, Veenstra TD, Panisko EA, Conrads TP, Angell NH, Smith RD (2002) Phosphoprotein isotope-coded affinity tags: application to the enrichment and identification
34.
35.
36.
37.
38.
39.
40. 41.
42.
43.
44.
45.
of low-abundance phosphoproteins. Anal Chem 74:607–616 Ficarro SB, McCleland ML, Stukenberg PT, Burke DJ, Ross MM, Shabanowitz J, Hunt DF, White FM (2002) Phosphoproteome analysis by mass spectrometry and its application to Saccharomyces cerevisiae. Nat Biotechnol 20:301–305 Cohen P (2001) The role of protein phosphorylation in human health and disease. The Sir Hans Krebs Medal Lecture. Eur J Biochem 268:5001–5010 Olsen JV, Blagoev B, Gnad F, Macek B, Kumar C, Mortensen P, Mann M (2006) Global, in vivo, and site-specific phosphorylation dynamics in signaling networks. Cell 127:635–648 Nita-Lazar A, Saito-Benz H, White FM (2008) Quantitative phosphoproteomics by mass spectrometry: past, present, and future. Proteomics 8:4433–4443 Yu Li-R, Issaq HJ, Veenstra TD (2007) Phosphoproteomics for the discovery of kinases as cancer biomarkers and drug targets. Proteomics Clin Appl 1:1042–1057 Mirzaei H, Mcbee JK, Watts J, Aebersold R (2007) Comparative evaluation of current peptide production platforms used in absolute quantification in proteomics. Mol Cell Proteomics 7:813–823 Wilson DS, Nock S (2003) Recent developments in protein microarray technology. Angew Chem Int Ed Engl 42:494–500 Calvert VS, Tang Y, Boveia V, Wulfkuhle J, Schutz-Geschwender A, Olive DM, Liotta LA, Petricoin EF (2004) Development of multiplexed protein profiling and detection using near infrared detection of reverse phase protein microarrays. Clin Proteomics J 1:81–89 Geho DH, Lahar N, Ferrari M, Petricoin EF, Liotta LA (2004) Opportunities for nanotechnology-based innovation in tissue proteomics. Biomed Microdevices 6:2319 Zhu H, Bilgin M, Bangham R, Hall D, Casamayor A, Bertone P, Lan N, Jansen R, Bidlingmaier S, Houfek T, Mitchell T, Miller P, Dean RA, Gerstein M, Snyder M (2001) Global analysis of protein activities using proteome chips. Science 293:2101–2105 Espina V, Wulfkuhle JD, Calvert VS, VanMeter A, Zhou W, Coukos G, Geho DH, Petricoin EF, Liotta LA (2006) Laser capture microdissection. Nat Protoc 1:586–603 Grubb RL, Calvert VS, Wulfkuhle JD, Paweletz CP, Linehan WM, Phillips JL, Chuaqui R, Valasco A, Gillespie J, Emmert-Buck M, Liotta LA, Petricoin EF (2003) Signal
Proteomics and Systems Biology: Application in Drug Discovery and Development
46.
47.
48.
49.
50.
51. 52. 53.
54.
55.
pathway profiling of prostate cancer using reverse phase protein microarrays. Proteomics 3:2142–2146 Tolcher AW, Reyno L, Venner PM, Ernst SD, Moore M, Geary RS, Chi K, Hall S, Walsh W, Dorr A, Eisenhauer E (2002) A randomized phase II and pharmacokinetic study of the antisense oligonucleotides ISIS 3521 and ISIS 5132 inpatients with hormone-refractor y prostate cancer. Clin Cancer Res 8: 2530–2535 Paweletz CP, Charboneau L, Bichsel VE, Simone NL, Chen T, Gillespie JW, EmmertBuck MR, Roth MJ, Petricoin EF III, Liotta LA (2001) Reverse phase protein microarrays which capture disease progression show activation of pro-survival pathways at the cancer invasion front. Oncogene 20:1981–1989 Wulfkuhle JD, Aquino JA, Calvert VS, Fishman DA, Coukos G, Liotta LA, Petricoin EF (2003) Signal pathway profiling of ovarian cancer from human tissue specimens using reverse-phase protein microarrays. Proteomics 3:2085–2090 Zha H, Raffeld M, Charboneau L, Pittaluga S, Kwak LW, Petricoin E III, Liotta LA, Jaffe ES (2004) Similarities of prosurvival signals in Bcl-2-positive and Bcl-2-negative follicular lymphomas identified by reverse phase protein microarray. Lab Invest 84:235–244 Sheehan KM, Calvert VS, Kay EW, Lu Y, Fishman D, Espina V, Aquino J, Speer R, Araujo R, Mills GB, Liotta LA, Petricoin EF, Wulfkuhle JD (2005) Use of reverse phase protein microarrays and reference standard development for molecular network analysis of metastatic ovarian carcinoma. Mol Cell Proteomics 4:346–355 Hanash S (2003) Disease proteomics. Nature 422:226–232 Tyers M, Mann M (2003) From genomics to proteomics. Nature 422:193–197 Agaton C, Galli J, Hoiden Guthenberg I, Janzon L, Hansson M, Asplund A, Brundell E, Lindberg S, Ruthberg I, Wester K, Wurtz D, Hoog C, Lundeberg J, Stahl S, Ponten F, Uhlen M (2003) Affinity proteomics for systematic protein profiling of chromosome 21 gene products in human tissues. Mol Cell Proteomics 2:405–414 Haab BB, Paulovich AG, Anderson NL, Clark AM, Downing GJ, Hermjakob H, Labaer J, Uhlen M (2006) A reagent resource to identify proteins and peptides of interest for the cancer community: a workshop report. Mol Cell Proteomics 5:1996–2007 Vorderwulbecke S, Cleverly S, Weinberger SR, Wiesner A (2005) Protein quantification
56. 57.
58.
59.
60. 61. 62.
63.
64.
65.
66.
27
by SELDI-TOF-MS based ProteinChip® system. Nat Methods 2:393–395 Wiesner A (2004) Detection of tumor markers with ProteinChip technology. Curr Pharm Biotechnol 5:45–67 Amaar YG, Thompson GR, Linkhart TA, Chen ST, Baylink DJ, Mohan S (2002) Insulin-like growth factor-binding protein 5 (IGFBP-5) interacts with a four and a half LIM protein 2 (FHL2). J Biol Chem 277:12053–12060 Boyle MDP, Romer TG, Meeker AK, Sledjeski DD (2001) Use of surface-enhanced laser desorption ionization protein chip system to analyze streptococcal exotoxin B activity secreted by Streptococcus pyogenes. J Microbiol Methods 46:87–97 Weinberger SR, Boschetti E, Santambien P, Brenac V (2002) Surface-enhanced laser desorption-ionization retentate chromatography mass spectrometry (SELDI-RC-MS): a new method for rapid development of process chromatography conditions. J Chromatogr B Analyt Technol Biomed Life Sci 782:307–316 Grunhage F, Fischer HP, Dauerbruch T, Reichel C (2003) Drug- and toxin-induced hepatotoxicity. Z Gastroenterol 41:565–578 Rej R (1989) Aminotransferases in disease. Clin Lab Med 9:667–687 Anderson NL, Taylor J, Hofmann JP, EsquerBlasco R, Swift S, Anderson NG (1996) Simultaneous measurement of hundreds of liver proteins: application in assessment of liver function. Toxicol Pathol 24:72–76 Cunningham ML, Pippin LL, Anderson NL, Wenk ML (1995) The hepatocarcinogen methapyrilene but not the analog pyrilamine induces sustained hepatocellular replication and protein alterations in F344 rats in a 13-week feed study. Toxicol Appl Pharmacol 131:216–223 Arce A, Aicher L, Wahl D, Anderson NL, Meheus L, Raymackers J, Cordier A, Steiner S (1998) Changes in the liver protein pattern of female Wistar rats treated with the hypoglycemic agent SDZ PGU 693. Life Sci 63:2243–2250 Meneses-Lorente G, Guest PC, Lawrence J, Muniappa N, Knowles MR, Skynner HA, Salim K, Cristea I, Mortishire-Smith R, Gaskell SJ, Watt A (2004) A proteomic investigation of drug-induced steatosis in rat liver. Chem Res Toxicol 17:605–612 Meneses-Lorente G, Watt A, Salim K, Gaskell SJ, Muniappa N, Lawrence J, Guest PC (2006) Identification of early proteomic markers for hepatic steatosis. Chem Res Toxicol 19:986–998
28
Chakravarti, Mallik, and Chakravarti
67. Gao J, Ann Garulacan L, Storm SM, Hefta SA, Opiteck GJ, Lin JH, Moulin F, Dambach DM (2004) Identification of in vitro protein biomarkers of idiosyncratic liver toxicity. Toxicol In Vitro 18:533–541 68. Goodsaid FM (2004) Identification and measurement of genomic biomarkers of nephrotoxicity. J Pharmacol Toxicol Methods 49:183–186 69. Han WK, Bailly V, Abichandani R, Thadhani R, Bonventre JV (2002) Kidney Injury Molecule-1 (KIM-1): a novel biomarker for human renal proximal tubule injury. Kidney Int 62:237–244 70. Bandara LR, Kelly MD, Lock EA, Kennedy S (2003) A correlation between a proteomic evaluation and conventional measurements in the assessment of renal proximal tubular toxicity. Toxicol Sci 73:195–206 71. Steiner S, Aicher L, Raymackers J, Meheus L, Esquer-Blasco R, Anderson NL, Cordier A
72.
73.
74.
75.
(1996) Cyclosporine A decreases the protein level of the calcium-binding protein calbindin-D 28kDa in rat kidney. Biochem Pharmacol 51:253–258 Aicher L, Wahl D, Arce A, Grenet O, Steiner S (1998) New insights into cyclosporine A nephrotoxicity by proteome analysis. Electrophoresis 19:1998–2003 Gupta N, Law A, Poddar R, Louie M, Ray A, Chakravarti DN (2005) Toxicoproteomics: applications in drug development. Curr Proteomics 2:97–101 Amacher DE (2010) The discovery and development of proteomic safety biomarkers for the detection of drug induced liver toxicity. Toxicol Appl Pharmacol 245:134–142 George J, Singh R, Mahmood Z, Shukla Y (2010) Toxicoproteomics: new paradigms in toxicology research. Toxicol Mech Methods [Epub ahead of print]
Chapter 2 Systems Biology Approaches and Tools for Analysis of Interactomes and Multi-target Drugs André Schrattenholz, Karlfried Groebe, and Vukic Soskic Abstract Systems biology is essentially a proteomic and epigenetic exercise because the relatively condensed information of genomes unfolds on the level of proteins. The flexibility of cellular architectures is not only mediated by a dazzling number of proteinaceous species but moreover by the kinetics of their molecular changes: The time scales of posttranslational modifications range from milliseconds to years. The genetic framework of an organism only provides the blue print of protein embodiments which are constantly shaped by external input. Indeed, posttranslational modifications of proteins represent the scope and velocity of these inputs and fulfil the requirements of integration of external spatiotemporal signal transduction inside an organism. The optimization of biochemical networks for this type of information processing and storage results in chemically extremely fine tuned molecular entities. The huge dynamic range of concentrations, the chemical diversity and the necessity of synchronisation of complex protein expression patterns pose the major challenge of systemic analysis of biological models. One further message is that many of the key reactions in living systems are essentially based on interactions of moderate affinities and moderate selectivities. This principle is responsible for the enormous flexibility and redundancy of cellular circuitries. In complex disorders such as cancer or neurodegenerative diseases, which initially appear to be rooted in relatively subtle dysfunctions of multimodal physiologic pathways, drug discovery programs based on the concept of high affinity/high specificity compounds (“one-target, one-disease”), which has been dominating the pharmaceutical industry for a long time, increasingly turn out to be unsuccessful. Despite improvements in rational drug design and high throughput screening methods, the number of novel, single-target drugs fell much behind expectations during the past decade, and the treatment of “complex diseases” remains a most pressing medical need. Currently, a change of paradigm can be observed with regard to a new interest in agents that modulate multiple targets simultaneously, essentially “dirty drugs.” Targeting cellular function as a system rather than on the level of the single target, significantly increases the size of the drugable proteome and is expected to introduce novel classes of multi-target drugs with fewer adverse effects and toxicity. Multiple target approaches have recently been used to design medications against atherosclerosis, cancer, depression, psychosis and neurodegenerative diseases. A focussed approach towards “systemic” drugs will certainly require the development of novel computational and mathematical concepts for appropriate modelling of complex data. But the key is the extraction of relevant molecular information from biological systems by implementing rigid statistical procedures to differential proteomic analytics. Key words: Network, Interactome, Multi-target drugs, Proteomics, Genomics, Systems biology, Posttranslational modification
Qing Yan (ed.), Systems Biology in Drug Discovery and Development: Methods and Protocols, Methods in Molecular Biology, vol. 662, DOI 10.1007/978-1-60761-800-3_2, © Springer Science+Business Media, LLC 2010
29
30
Schrattenholz, Groebe, and Soskic
1. Genetic Interaction Network Analysis
Large-scale genome sequencing efforts have indicated an enormous scope of interactions, albeit only on the level of nucleic acids (1–3). Most of the insights into genetic interactions and networks come from studies using the yeast Saccharomyces cerevisiae (4, 5) but also more recently from multicellular organisms such as Caenorhabditis elegans (6) and mammalian cancer cell lines (7–9). The first genetic interaction maps in yeast were based on “synthetic lethal” interactions causing cell death by a combination of two gene deletions which by themselves have no deleterious effects. This type of phenomenon is also named “epistasis,” and is commonly used to define “genetic interactions” in statistical terms (10). In eukaryotic organisms and cell lines, the recent development of RNAi libraries has enabled systematic genetic studies by using arrayed and pooled screens (8, 11). A general understanding of the topology of genetic interaction networks in yeast has a wider importance, because similar principles are expected to underlie the relationship between genotype and phenotype in higher eukaryotic species. In terms of human disease, numerous modifiers and enhancers contribute to the complexity of phenotypes on a background of genetic, epigenetic and posttranslational stochasticity, but the structural and temporal determinants of the underlying networks remain unknown. Thus, mapping genetic networks in model organisms, such as yeast, supplies a framework for more complex systems, but next to structural interactions, temporal and spatial resolution of cellular sub compartments on the level of fast posttranslational events of proteins constitutes an important next level of modelling attempts. Physical-interaction maps, generated by large-scale twohybrid screening (12, 13) or affinity purification schemes followed by mass spectrometry (14–16), thus only provide an indirect link between genes and functional protein complexes functioning on time scales completely different as compared to nucleic acid turn over. So, next to valuable, but limited physical information (mostly taking into account structures related to the mere amino acid backbone of proteins, devoid of structural and kinetic consequences of posttranslational modifications), the genetic interaction map provides only limited functional information, largely identifying gene products that operate in functionally related pathways. Bluntly, the genetic information cannot tell when, where and how long protein partners will interact and what consequences result from very fast chemical changes, for example, oxidation, phosphorylation or proteolysis. However, one significant finding made by systematic deletion studies in yeast showed that most eukaryotic genes are dispensable
Systems Biology Approaches and Tools for Analysis of Interactomes
31
for viability, that is, there is an overwhelming degree of redundancy. Only about 20% of Saccharomyces cerevisiae genes are essential for haploid cells grown under laboratory conditions (3, 17). On the other hand, topological analysis of the yeast genetic interaction network predicts roughly 200,000 synthetic lethal interactions in the global yeast genetic network. Extrapolated to humans, this translates to a huge number of interactions with a specific phenotype potentially influenced by hundreds of different gene combinations (18). Only in a minority of cases, interactions can be deduced from genes or the directly related amino acid backbone information of proteins. In the vast majority of cases, low affinity interactions in highly redundant signalling pathways on the level of fast chemical changes at amino acid side chains play a key role. However, the genetic framework provides a lead idea about rough organizational principles, and here phylogenetic principles may offer important clues (19). Drug interactions happen to take place at higher levels of deconvolution or the unfolding of the genetic information: subtle structural alterations by small molecules attenuating downstream effects of gene mutations as rapid and reversible modulators of related protein activity and structure. The use of such chemical probes on a genome-wide scale is called “chemical genomics” and has been originally used in yeast but also more recently in metazoan cell lineages. The underlying idea, that deletion of a gene encoding the target of an inhibitory compound should cause cellular effects that are similar to the inhibition of the target by drug treatment is of course simplistic in the light of the arguments developed above, but in a few cases helps shed light on major components of pathways. In one study (20), the chemical– genetic profiles of five different compounds were found to be reasonably similar to the genetic-interaction profiles of the target gene or genes in the target pathway. In another example, synthetic lethal genetic interactions were hopeful for the identification of compounds that target specific pathways and selectively kill cells with defined mutant genotypes in cancer models (21). Using a metazoan retroviral expression system, Burkard et al. (22) identified pleiotropic functions of polo-like kinases 1 throughout mitosis and cytokinesis and demonstrated some usefulness of chemical genetics in dissecting these complex but shortlived events within human cells. A similar strategy was used to replace cyclin-dependent kinase 7, the cyclin-dependent kinaseactivating kinase, with an analogue sensitive version in human cells (23). However, given the large number of kinases and the highly redundant biochemical pathways involved in mitosis and cytokinesis, the further application of chemical genetic analysis of the metazoan cell division cycle has to be further validated. In general, genetic approaches have provided a tremendous amount of largely descriptive information and most often lack to
32
Schrattenholz, Groebe, and Soskic
adequately represent the relevant proteomic level, where we observe a considerable inflation of numbers of molecular species due to posttranslational modifications and moreover intricate kinetic features which require tight analytical synchronisation due to huge dynamic ranges of concentration differences and time scales of changes, all this in frameworks of a huge degree of biological stochasticity (24).
2. Multi-target Drugs There is a broad understanding that the success of target and lead structure identification, optimization, pre-clinical validation and clinical development has always depended on a thorough understanding of underlying biological mechanisms in disease and corresponding treatment. The reality of current pharmaceuticals out in the market, however, includes a significant segment of successful drugs which have low selectivities and affinities and multiple targets (24) where the exact modes of action are far from clear. Given, for example, the complex and polyetiological molecular biological background of central nervous system disorders, functional genomic and in particular proteomic profiling of efficacy models (25, 26) increasingly contribute to drug development (27–30). The enormous functional flexibility of many protein targets results in crosstalk and pleiotropy, generating combinatorial effects and cascades of non-linear cellular functions. A decade ago and based on the hypothesis that 100,000 open reading frames for proteins exist in the human genome, Drews and Ryser proposed roughly 500 molecular targets interacting with marketed drugs (31). Today, after realising that there might be only about 20,000 genes in Homo sapiens on the one hand, but that due to posttranslational modifications of proteins on the other hand, there might exist millions of distinct molecular species, the situation is still unsettled. If a target is considered to be “a molecular structure (chemically definable by at least a molecular mass)” that will undergo a specific interaction with given chemicals, we might end up with a bigger number of potential targets, may be 1,000 (32). However, this type of definition as so-called minitargets based on one major and predominant biochemical interaction may prove to be too reductionist. In fact, most drugs appear to interact with several targets over a range of affinities. Therefore, every substance has a unique “effect-space” in biological homeostasis, which usually is far out of equilibrium and energetically highly “charged.” The subtleties imposed on drug/target interactions by chemical modifications of amino acid side chains of proteins like phosphorylation, methylation, and ADP-ribosylation, as well as proteolytic processing, offer new chances of defining
Systems Biology Approaches and Tools for Analysis of Interactomes
33
drugable substructures beyond the mere amino acid backbone of translated genetic information (24). Living dynamic systems react to external input on time scales starting in the nanosecond range with consequences potentially lasting years, but research tools used in drug development are generally limited to study targets as static objects (for example, 3D protein structures). A relevant “target” often exhibits varying degrees of freedom depending upon physical and chemical conditions of cellular micro-environments. The sum of these interactions will give as an output an effect profile. So far we are unable to adequately address these so-called macro-targets. Current high through-put in vitro screening techniques appear to identify such targets only by serendipity. Targeting net effects rather than direct high affinity chemical interactions will require novel concepts, integrating genetic and proteomic information about redundant, pleiotropic and recurrent biological signalling (33, 34). In this regard, concepts like gene expression state spaces and attractors have been introduced which try to provide a mathematical and molecular basis for “epigenetic landscapes” (35). Dynamic transient behaviours of cellular reactions have been discussed in terms of chaotic itinerancy (36), and in particular the aspect of multi-degree-of freedom reactions has led to the analysis of high-dimensional variance data. There are approaches to reveal and interpret the structure in such data spaces or “landscapes” or signatures, essentially by geometrical and correlational procedures (37). The application of these concepts to human diseases and drug development requires general approaches for the quantification of synergies of combinatorial effects (multiple modes of action of a single drug or a combination of drugs). The relevance of nonlinear dose response surfaces in complex biological systems has been shown in HIV and cancer chemotherapy by drug combination experiments. In this context, the definition of exact sites of drug interactions with protein target domains in vitro is the first important step. However, related screening usually employs recombinant proteins devoid of posttranslational modifications, ignoring the considerable flexibility in protein structure which can shape signalling decisively. The attractiveness of this reductionist approach, which has set the standards in pharmaceutical research for decades, lies in the fact that high through-put can be achieved in screening, albeit as we have been suspecting for a while, at the cost of relevance. Today, we face new challenges moving forward to develop rationales for the analysis of drugs with multiple modes of action, often with moderate affinities and poor selectivities on the background of a daunting biological complexity. Indeed, new scientific paradigms can be expected to emerge in particular with regard to mathematical treatment and modelling (38–41).
34
Schrattenholz, Groebe, and Soskic
Healthy biological signalling is always far out of equilibrium, consuming constantly high amounts of energy. Control of energy homeostasis is maintained by a first level of key metabolites or second messengers including intracellular calcium concentrations, ATP, and NADH/H+. The “energy” status of a cellular system is partly integrated by two posttranslational modifications consuming energy-rich substrates: phosphorylation and ADP-ribosylation, which are commonly found to be regulated on very fast time scales in almost every major signalling pathway. In particular, the activation of most types of receptors at some point involves phosphorylation cascades. Cellular stress on the other hand converges on the intrinsic or mitochondrial apoptotic pathway of cell death. Many calcium-driven pathologies, apoptosis, ATP, energy, reactive oxygen species (ROS), and cell death converge at the level of mitochondria (42–45). These disequilibria result in simultaneous effects of multitarget drugs across several biochemical pathways or in different organs. Because of the nature of living systems, the net result will not be linearly deducible from single effects, but will be nonlinear, with particular “effect spaces or surfaces” with a continuum between beneficial drug effects and toxic side reactions. For drug combinations, this is even more complicated. A mechanism-based simulation of pharmacodynamic drug–drug interactions was described recently (46). For this reason, the principle of blocking a single pharmacological target with high potency has been seducing because it appears to minimize side effects coming with less specific, multi-target drugs. However, in complex diseases like cancer or neurodegenerative disorders, we observe a manifold of genetic and epigenetic contributions to gradually and accumulating cellular damage. Age is one of the most common risk factors (47) and mitochondrial dysfunction plays a central role. The mode of action of current medications in, for example, the treatment of Alzheimer’s disease, like acetylcholine esterase inhibition or NMDA receptor channel blockade only remotely (if at all) correspond to firmly established genetic risk factors such as amyloid precursor protein, presenilins, neuregulin-1 or ApoE4 (48–52). There is an increasing awareness that mitochondrial and epior postgenetic mechanisms like the assembly and processing of receptors and other membrane proteins in cholesterol-rich “rafts” might have a key role in the manifestation of genetic risks (33, 53, 54). Even in the case of successful drugs like Genentech’s breast cancer treatment Herceptin and Novartis’ leukaemia drug Gleevec, each of which targets a specific genetic mutation, therapeutic effects are not stable. Eventually, these drugs become inactive in many patients because of gradually increasing resistance of cancer cells. Some of these escape mechanisms are systemic
Systems Biology Approaches and Tools for Analysis of Interactomes
35
responses far beyond single-target mechanisms, as the emerging example of cancer stem cells shows, where a minority cell population evolves into something like an alien organism inside the mother organism under the pressure of the immune defence (55–58). The example shows that cancer is not just the result of strayed genes, but merely starts there, proceeding into an ongoing evolutionary battle for survival against host defence. Mutations in at least 189 genes have recently been described in human breast and colorectal cancers (59). These findings support concerns that monofunctional drugs will not cure most of common carcinomas. The number of genes relevant to carcinogenesis in general could be much higher (60), because mutations in non-coding regulatory regions of coding genes may contribute as well. Given the relatively small number of genes, we have to realize that advanced cancer represents a complete systemic reorganisation, with cancer cells often being more alien to the organism, in which they reside, than a closely related species. Curing cancer will require understanding (and interrupting) these complex escape mechanisms. Eventually, cocktails of drugs are more likely than single magic bullets (61, 62). For example, it was shown that the inhibitors of the arachidonic acid pathway and peroxisome proliferator-activated receptor ligands have superadditive effects on lung cancer growth inhibition (63). When given together they are much more efficient in killing cancer cells than each of them given alone. Synthetic oleanane triterpenoids and retinoids, two new classes of multifunctional drugs (64) are neither conventional cytotoxic agents nor are they monofunctional drugs that uniquely target single steps in signal transduction pathways. They have unique molecular and cellular mechanisms of action and might prove to be synergistic with standard anticancer treatments. An opposite example is Vioxx: recent problems with this highly selective cyclooxygenase-2-inhibitor are considered to be due to its extreme selectivity which seems to tip the balance of pro- and antithrombotic mediators in an unfavourable way (65). Also the history of antipsychotics shows that “dirty” or “promiscuous” drugs might be increasingly important in the future (66, 67). The first generation of antipsychotic drugs; discovered about 50 years ago (e.g., chlorpromazine), were described as selective D2 dopamine antagonists, and today are known as “typical antipsychotics” (68, 69). These drugs produce a host of adverse side effects, without having any effect on the negative symptoms of schizophrenia. Later “atypical antipsychotics” were introduced (70) which cause significantly less extrapyramidal side effects. On the molecular level, the difference between the two classes can be attributed to a different binding profile to D2 dopamine and 5-HT2A serotonin receptors. The second generation of atypical antipsychotics
36
Schrattenholz, Groebe, and Soskic
(also known as “dopamine-serotonin system stabilizers”) such as aripiprazole, offered further advantage, due to an improved efficacy in treating the negative symptoms of schizophrenia and a decreased incidence and severity of central and peripheral side effects (70). Receptor binding studies revealed that the secondgeneration of atypical antipsychotics are partial agonists of D2 and 5-HT1A receptors, and antagonists of 5-HT2A receptors but also a number of other targets (24). Another example is (−)-epigallocatechin gallate (EGCG), the major constituent of green tea, causing induction of apoptosis and cell cycle arrest in many types of cancer cells without affecting normal cells (71, 72). Apoptosis is a highly ordered protective mechanism with clear mitochondrial connotations through which unwanted or damaged cells are eliminated from the system and EGCG has been shown to affect a number of targets (24). A number of subsequent studies have shown similar effects of other dietary constituents (73). Three strategies are available to the designers of multitarget therapies. The first strategy is to prescribe multiple individual medications. The drawback is patient compliance and the danger of drug–drug interactions. To overcome this problem, a second strategy is the development of multi component medications that contain two or more active ingredients formulated in the same delivery vehicle, such as a single pill. However, due to significant differences in pharmacokinetics, metabolism and bioavailability, the galenics and formulation of drug combinations are no trivial problems. Further, two drugs that are safe when dosed individually cannot be assumed to be safe in combination. Another drawback using drug combinations is that the target space of the current pharmacopeia is limited to approximately 1,200 FDAapproved drugs. The third strategy is to design a single compound acting on multiple targets (74, 75). Dosing with a single compound may have advantages over a drug combination in terms of equitable pharmacokinetics and biodistribution.
3. Analysis of Modes of Action of Multi-target Drugs Using a Network Approach
The rational design of drugs with multiple targets necessitates the molecular analysis of their modes of action in appropriate cellular models. Cellular networks form four major classes: metabolic pathways, gene regulation networks, signal transduction networks and protein interaction networks (76). In a living cell, at least three, possibly all four, of these networks interact: requiring more holistic strategies of investigation. In the network concept, the cell is perceived as a set of interacting elements, which are connected by links. Links have a weight, which characterizes their
Systems Biology Approaches and Tools for Analysis of Interactomes
37
strength defined for example by affinity or propensity. Links may also be directed links, when one of the elements has a larger influence to the other. Interacting molecules are considered as elements in these networks, and their interactions form the weighted, but not necessarily directed links. Alternatively, we may also envision directed links as representations of signalling or metabolic processes of the functional networks in the cell. Cellular networks contain hubs, that is, elements, which have a large number of neighbours. These networks can be dissected to overlapping modules, which are supposed to form hierarchical communities (77, 78). Computational cellular network models offer a lot of possibilities to identify nodal elements as potential drug targets. In recent years, several experimental and modelling approaches have been attempted to identify targets in a network and systems biology context (79–83). The major analytical challenges result from the complexity of protein expression in terms of linear dynamic range, the number of biochemical species and the dynamics of posttranslational modifications. The time scales range from fractions of seconds (e.g., phosphorylation, regulation by proteolytic cleavage) to many days, months or even longer (e.g., permanent modification of synapses during learning, initiated by an activity-dependent phosphorylation cascade). If we accept the notion, that multi-target drugs have their effects on the level of effect spaces defined mainly by protein isoforms, then the first prerequisite is an apparently simple one: statistically significant differential control of protein expression in appropriately treated biological samples. Recently, it became clear that this is not a trivial task due to huge ranges of abundances, kinetics and chemical diversity of proteins (84). Dealing with convoluted biological networks of extreme flexibility and redundancy has even let to consider “game theory approaches” as useful for the simplification of complex sets of nonequilibrium conditions by the introduction of “multi-target drug design games” (85). Simple topological network models will only suffice to provide preliminary insight (Fig. 1). In such networks, the modelling of pharmacological modes of action is attempted by representing general elements of the network in a sense of functionally defined target classes (proteins, RNA- or DNA-sequences). Connecting links are meant to represent known interactions of single entities within the cell. However, since thus far only amino acid backbone information of proteins (essentially amounting to the genetic information available) is included, these concepts have not delivered validated predictions yet. Drug-induced effects upon single targets are in an initial step modelled by the elimination of all interactions at the representing element (complete knockout), which essentially ignores any degree of redundancy and pleiotropy. The much more frequent and relevant partial inactivation of drug targets has been modelled in different ways: either by knocking
38
Schrattenholz, Groebe, and Soskic
Fig. 1. Topological molecular networks of the yeast cell interactome, adapted from Ho et al. (98). (a) Presentation of the full Filtered Yeast Interactome data set. Differences in node size are due to the 3-D layout rendering nodes projected towards the viewer that are larger than those that are projected away from the viewer and are not indicative of any differences in proteomic attributes. (b) Protein interactions and associated data in the biggest connected component of the yeast nuclear interactome. Yellow RNA metabolism; orange organelle organization and biogenesis; light blue protein biosynthesis; dark blue cell cycle; green transcription; red process unknown. (c) Proteins colored by localization. Blue nucleus; light blue nucleolus; green mitochondrion; gold cytoplasm; red unknown. (d) Textual annotation of a portion of the nuclear interactome, showing nodes labelled by gene name. (e) The same nodes annotated for post-translational modification: P is phosphorylation and A is acetylation. The number associated with the modification type indicates the number of known sites per protein.
out complete parts of the interactive options of a given protein or by attenuating more moderately selected interactions of a protein. In simplified models, the “attack” on a network, such as the genetic regulatory networks of Escherichia coli (86, 87) or Saccharomyces cerevisiae (88, 89), it was reported that multiple but partial attacks on carefully selected targets are more efficient than the complete knockout of a single, though equally well selected, target (90). The largest effect to the E. coli regulatory network is reached by removing one element with 72 connections. However, the same effect is achieved by partial inactivation of three to five network elements, despite the observation, that the number of affected interactions does not increase (90). This example was based on network topology and applied to antimicrobial drugs, where network damage corresponds well to desired drug effects. Bacterial models are especially well suited for modelling attempts because of the nearly complete absence of posttranslational modifications in prokaryotes.
Systems Biology Approaches and Tools for Analysis of Interactomes
39
In principle, this can be useful for novel concepts about the development of “multidimensional” drugs in much more complex mammalian models. We have to consider why this is so; on the one hand, microbes are far less complex than mammalian cells due to lack of posttranslational modifications. On the other hand, it seems that during the course of evolution, the complexity observed in higher biological systems is just representing manifold feedback variations of relatively clear-cut “original” and basic enzymatic and structural biological themes; many of them have in principle been invented by prokaryotes (19). The major evolutionary reason for this inflation of molecular species by posttranslational modifications of proteins appears to be the resulting redundancy, crosstalk and compensation and thus an enormous functional flexibility. By small changes of activity-dependent parameters, for example, intracellular calcium, qualitatively completely different, even opposing functional reactions can be elicited. Moreover, redundancy is one of the major cellular insurance mechanisms against jeopardy of the whole system by single detrimental effectors. We simply observe cellular stress-management and it may be not surprising that modes of action of drugs obey similar principles. In many cases the receptors, enzymes and other biomolecules involved in the networks of stress management are the very same responsible for physiological signal transduction (24). For the rational design of multi-target drugs, much more detailed information about kinetics, localisational and biochemical dynamics of specific signalling, metabolic and transcriptional pathways would be required. And although observing a shift of paradigm towards highly efficient low affinity, multi-target drugs, the challenges resulting from the complexity of systems require novel approaches and tools in analytical, computational and modelling efforts (91–93). So far, some progress has been made by opportunistic approaches (25, 26, 76), but a rationale for systematic screening of multiple-target compounds, with moderate affinities and selectivities or even appropriate mixtures of drugs is still in the making. The main difficulty is adequate representation and synchronisation of fast chemical changes and resulting effects on activity and structure of biomolecules.
4. Analysis of Action of Multitarget Drugs Using Systems Biology Tools
The screening of compounds and compound libraries for this type of mechanism requires an intricate combination of advanced technologies: genomic and proteomic analyses correlated directly to biochemical and/or cell biological in vitro assays might be most appropriate and fastest. Before multimodal drug responses
40
Schrattenholz, Groebe, and Soskic
can be validated in vivo, functional in vitro models for human or animal tissues, which simultaneously can be submitted to molecular analysis or genetic manipulation, are mandatory. Models like embryonic stem cells and their organ-specific derivatives can be submitted to functional kinetic read-outs by fluorescent techniques, and at the same time being correlated directly to molecular snapshots of biomarker signatures obtained by microarray or differential proteomic techniques (94, 95). The complexity of biological systems thus also requires convoluted analytical procedures, in particular, adequate differential procedures for quantitative kinetic measurements and the subsequent independent validation of protein biomarkers, and there is no highthroughput method or screening readily at hand (19, 84, 96). 4.1. Differential Quantification of Protein Biomarker Signatures and the Study of Interactomes
Proteomics provides the parallel analysis of large numbers of proteins (19, 96). However, most proteomic technologies are relatively poor in exactly measuring protein abundances over the large linear dynamic range of typical biological samples (24). There are essentially only three methods for a reliable quantification of differential proteins in complex mixtures: labelling with stable or radioactive isotopes or with fluorescent dyes, usually Cy-dyes. The statistical treatment of differential quantities measured for different sample conditions is crucial and depends on the dynamic range and resolution of the separation method. The underlying statistical concepts have been discussed in depth recently (24). In essence, raw differential quantities of labelled and separated proteins have to be normalized, the reproducibility of the detection method and potential labelling bias have to be controlled by statistical procedures, like Bland-Altman and MA-plots. Subsequently, repeat measurements of inversely labelled samples need to be statistically analysed in terms of significance of amplitude differences (Volcano plots, showing the relation between p-values and intensity differences). These procedures serve to identify robust protein biomarker candidates from complex mixtures of proteins and surrogate patterns of multidimensional LC-MS or 2D-PAGE analysis (84). These considerations clearly point to the dynamic range of the detection method and resolution of separation technique as the most crucial parameters for finding valid biomarkers in complex biological or clinical samples (84, 97). The increasingly recognized role of posttranslational modifications justifies laborious “wet” proteomic approaches, because only in this way relevant molecular content information can be generated from native biological material. Given the functional significance of minute structural modifications of proteins in multidimensional interactomes, it is frankly questionable which artifactual effects might result from the interpretation of non-linear data by the introduction of recombinant affinity or fluorescent
Systems Biology Approaches and Tools for Analysis of Interactomes
41
tags into interacting proteins. The essential failure of high-throughput array data or in silico approaches to contribute to novel modes of action and innovative drug design shows the necessity of quantitative proteomic data from appropriate native models before approaching a realistic modelling of drug effects. The degree of stochastic and non-linear behaviour in living systems is amazing and the complexity of biological systems as well. The codes of kinetically resolved and statistically significant protein signatures or molecular snap shots necessary for the screening of novel multifunctional drugs so far are beyond our understanding. However, we can use protein signature information for predictions. The mathematical concepts for appropriate modelling tools and the analytical procedures needed for protein signature-based prediction are currently in the making and under hot debate. The main caveat of current methodology which is impeding computational efforts towards an integrative treatment of protein interactions, protein abundances and localizations, is based in the intrinsic structural variability by posttranslational modifications and the fact that recombinant methods only distantly relate to and reflect biological reality (98). Protein turnover is largely contributing to protein signature snapshots and cannot be assessed by recombinant methods. Turnover of proteins after traditional and non-traditional ubiquitinylation, SUMO-ylation and subsequent degradation and regulation are increasingly recognized as major epi- or postgenetic principles (99–101). We see an increasing focus on the integration of quantitative “native” proteomic data, with novel concepts regarding the acquisition of quantitative biological data, their functional synchronisation, statistics, validation and modelling (102). 4.2. Data Storage and Integration, Modelling
As an example for current problems and approaches to data storage and modelling, the authors provide an outline for strategies in creating joint data repositories for complex data sets generated in the frame of a number of projects which are part of the European Community Sixth Framework Programme. These are ageing- or toxicology-related projects assembling genomic, transcriptomic, proteomic and functional data from a variety of models (MIMAGE; www.mimage.org; Reprotect; www.reprotect.eu; ESNATS; www. esnats.eu). Within MIMAGE, age-dependent protein maps and signalling pathways emerging from common traits in mitochondrial proteins isolated from five species in a variety of ageing models have provided a first glimpse on initial key events in fundamental ageing mechanisms. One example is the age-dependent oxidation of tryptophan-residues to N-formyl-kynurenine by ROS in SAMdependent O-methyltransferase (PaMTH1, in Podospora anserina, a fungal ageing model). The human homologue of PaMTH1, COMT plays a key role in the etiology of Parkinson’s disease.
42
Schrattenholz, Groebe, and Soskic
Parkinson’s disease is an age-dependent disease characterized by protein aggregation (a-synuclein-plaques) in dopaminergic neurons. Mitochondrial dysfunction has been identified as a major factor contributing to the onset of Parkinson’s disease. Also, COMT and dopamine metabolism are major therapeutic targets in schizophrenia (they are important in cognition). An integration of such a result into models of underlying signalling requires that data sets of age-dependent molecular and functional events, such as differential quantitative proteomic data (images, numbers, mass spectra), functional correlates (enzymatic activities, native multi protein complexes, calcium, electrochemical potentials, ATP), genomic information (longevity mutants, site-directed mutagenesis, knock-in-knock-out’s) and other biological background (medline) are first organized in an appropriate standardized fashion. Traditional proteomic differential analysis compares the proteomes of two experimental conditions to each other. With more advanced techniques, (19) a multitude of different conditions or experimental interventions can be compared at once. As a result, matched vectors of several hundred protein abundances (“proteomic inventories”) under a number of experimental conditions become available. This new quality of experimental data promotes a demand for joint analyses of several individual comparisons: –– How do experimental conditions relate to each other? –– Is the proteomic inventory of one condition similar to or largely different from the inventory of another one? –– Which conditions exhibit the most diverse proteomic inventory? –– Which combinations of conditions may be expected to represent extreme points of protein abundances? –– Can subsets of all protein spots be defined that show similar (or most different) abundances in certain (combinations of) experimental conditions? With these perspectives in mind, we suggest the following structure for data repository for advanced systems biology projects. 4.2.1. Data Repository Structure
When planning a systems biology data repository, several requirements are to be met: ●
●
●
The intrinsic structure of all data layers needs to be represented by the data storage format in a natural fashion. Data exchange with outside researchers and existing data storage solutions need to be easily possible. The effort necessary for setting up and maintaining the structure of the database solution must not be unreasonably high.
Systems Biology Approaches and Tools for Analysis of Interactomes ●
43
Data need to be made available in a format that is suitable to serve as input for system biological studies.
The last requirement calls for employing standardized data formats at least for data types of higher complexity. During the past years, a substantial effort has been spent on defining and agreeing on universally accepted minimum information requirements for the description of typical experimental procedures in various biological sub-fields. Examples of the results of such efforts are MIAME (Minimum Information about a Microarray Experiment (103)), MIBBI (Minimum Information for Biological and Biomedical Investigations (104)), MIFlowCyt (Minimum Information about a Flow Cytometry Experiment), MINI (Minimum Information about a Neuroscience Investigation), or MIAPE (Minimum Information about a Proteomics Experiment (105)). From verbose descriptions of minimum information required to describe certain types of experiments, abstract models of corresponding data structures have been developed. For example, the Functional Genomics Experiment Object Model (FuGE-OM (106)) arises from MIAME, or the mzData and analysisXML data models for capturing peak list information and representing mass spectrometry informatics data are (partial) materializations of MIAPE. Both models are advanced by the HUPO Proteomics Standards Initiative (107, 108) (http:// www.psidev.info/index.php?q=node/80). Especially in projects, where a large body of proteomic data is generated, the MIAPE standards have a particular value. To our experience, data repositories of the type sketched above require an integration of a number of existing database solutions to keep investments of man power and other costs in acceptable limits. We suggest combining two separate database frameworks specialized in storing different aspects of biological experimental result data. For the general description of the experimental conditions and for simply structured input and output data, the FuGE compliant “SYstems and Molecular Biology data and metadata Archive” (SyMBA) is used (http://symba.sourceforge.net/ SyMBA-Mged2008.pdf; developed by CISBAN, Centre for Integrated Systems Biology of Ageing and Nutrition, Newcastle University, http://www.cisban.ac.uk/). The data resulting from proteomic investigations, on the other hand, are stored in a modified “PRoteomics IDentifications database” (PRIDE) (109) (EBI, European Bioinformatics Institute; http://www.ebi.ac. uk/pride/) which is based on the mzData and analysisXML data models. Technically, the SyMBA installations are implemented as a Hibernate Software ToolKit (STK) encompassing a FuGEstructured relational backend database (based on PostgreSQL),
44
Schrattenholz, Groebe, and Soskic
a Java Object-Relational Persistence layer (Hibernate and Spring), a set of Java Object entity classes representing FuGE UML entities, and a set of Data Access Object (DAO) classes that facilitate and encapsulate access to entity classes. SyMBA employs FuGE (http://fuge.sf.net) as the core of its database structure, and provides means for using both MIBBI checklists (http://www.mibbi. org/), OBI controlled vocabulary (http://purl.obofoundry.org/ obo/obi). As a particular strength, SyMBA features highly elaborate built-in appliances for standardizing, structuring and storing experimental metadata for a range of omics experiments which include the following packages (106): ●
●
●
●
●
●
●
●
●
●
Audit (storing contacts, auditing and security settings for all objects), Description (to allow for additional annotations and free-text descriptions for all objects), Measurement (defines slots for providing atomic, Boolean, range and complex values with appropriate units, sourced from an ontology), Ontology (provides a mechanism for referencing external ontologies or terms from a controlled vocabulary), Protocol (represents a model of procedures, software, hardware and parameters and can define workflows by relating input and output materials and/or data to the protocols that act on them), Reference (enables external bibliographic or database references that can be applied to many objects across the FuGE model), ConceptualMolecule (captures database entries of biological molecules such as DNA, RNA or amino acid sequences and provides an extension point for other molecule types, such as metabolites or lipids), Data (sets the dimensions of data and storage matrices, or references to external data formats), Investigation (defines an overview of the investigation structure by capturing the overall design and the experimental variables and by providing associations to related data), Material (models material types such as organisms, samples or solutions. Materials are characterized by ontology terms or by the extension of the Material package).
By means of these packages, SyMBA offers data types for modelling experimental objects such as samples, protocols, instruments, or software modification histories. Finally, data describing specific input and output values of the respective experiments may be represented by defining suitable elements in FuGE’s Data Collection object.
Systems Biology Approaches and Tools for Analysis of Interactomes
45
PRIDE is a database for protein and peptide identifications that may potentially have been published in the literature, which will typically arise from specific species, tissues and subcellular locations, perhaps under specific disease conditions, and for which post-translational modifications may have been located on individual peptides (109). The PRIDE database definition and Java classes may be downloaded from EBI and built using Maven. Employing the Hibernate persistence framework, the code can be modified to accommodate additional fields and to allow for joining PRIDE and SyMBA database applications. In particular, it is necessary, ●
●
●
To include further fields in the protein identification table describing the quality of the identification, To define/supplement structures for the specification of the material which the identification originates from (electrophoresis gel and protein spot identifier), and To store the outcome of quantitative comparisons of protein abundances between two different biological conditions (typically two age groups).
The SyMBA and PRIDE parts can be connected as follows: The experimental details, in particular, the specific conditions entering each proteomic comparison are described in the SyMBA database. Each one of these comparisons is labelled by a unique identifier. Similarly, each result of such a comparison is labelled by the same unique identifier in the PRIDE part of the database. This allows formulating queries over both parts of the combined database in a unified fashion. For example, by matching these unique identifiers, all results of a given comparison in the SyMBA branch can be retrieved from the PRIDE branch, or the comparison details for a given result in the PRIDE branch can be retrieved from the SyMBA branch. Both parts of such a database may be generated from (more or less) abstract model descriptions via an automated process including MAVEN. All code is controlled using the Subversion version control system. Basic experimental data can be entered via web interfaces. More complex data may be uploaded as files (*.txt, *.doc, *.xls, *.csv, …). For proteomic data, suitable import routines have been programmed. Data may be accessed via versatile search interfaces which allow for, for example, finding all occurrences of a given protein in any one of the experiments in the data repository. More extensive data collections may be retrieved from the database via file download. 4.2.2. Employing Proteomic Data in Mathematical Modelling
Enzymes are organized in metabolic networks, and overall network function depends on the function of an individual enzyme in a highly intricate fashion. Therefore, it often is not at all obvious how activity- or age- (or generally input-) related changes in
46
Schrattenholz, Groebe, and Soskic
enzyme abundance (that have been diagnosed in the course of a differential proteomic analysis) affect network performance. To answer this type of questions, one can use a mathematical model of the respective metabolic pathways to study the effects of changes in abundance or activity of an individual enzyme on the entire network. For purposes of modelling metabolic pathways, a number of tools are publicly available: ●
●
●
The KEGG electronic pathway maps (110) (http://www. genome.ad.jp/kegg/pathway/map/map01100.html) give detailed information about the enzymes involved in metabolic networks. KEGG maps are editable and may be modified/annotated for presentation (111). However, KEGG does not provide for direct access to enzyme kinetic data. For retrieving numerical values for the enzyme kinetic constants, web-based database services are available which can be searched for the kinetic constants of an enzyme of interest, for example, “BRENDA – The Comprehensive Enzyme Information System” (112) (http://www.brendaenzymes.org/). Setting up the code for computing network behaviour is greatly facilitated by modelling frameworks like the “Systems Biology Workbench” (SBW) (113) (http://www.sys-bio. org/sbwWiki/doku.php?id=sysbio:sbw). The generated models are stored in systems biology markup language (SBML) (114) which may be imported into a variety of other modelling packages. The SBW framework includes the graphic network editor JDesigner for setting up metabolic network geometry and kinetics.
In the process of modelling a number of prerequisites need to be met: ●
●
●
The elements of the metabolic network need to be sufficiently well known. Sub-networks of the overall metabolic network need to be identified, which are largely self-contained but sufficiently comprehensive to generate biologically meaningful models. Realistic numerical values for the kinetic constants of all involved enzymes need to be acquired from the literature, or more easily from a web-based database service like BRENDA (112). However, great care must be taken in the factual choice of the numerical values because it is not uncommon that enzyme databases report functional parameters for exactly the same enzyme deviating by up to two orders of magnitude. Moreover, kinetics may differ greatly between isoforms of the same
Systems Biology Approaches and Tools for Analysis of Interactomes
47
enzyme, related species and even tissues. Therefore, it is generally not sufficient to just look up the values of interest in the database and plug them into the model. Rather, the exact conditions and methods of measurement need to be assessed from the original publication and adequately accounted for when fixing model parameters. This may also be achieved by statistical approaches (115). Moreover, it is a good idea to perform and take into account model runs for a number of sensibly selected parameter ranges. ●
●
●
●
Similarly as enzyme kinetic constants, enzyme volume-related activities need to be input in the models. For some enzymes, specific activities may be found in BRENDA. In other cases, data may be obtained from original investigations of enzyme kinetics in which material from animal tissues was used. Frequently, series of measured activities during the purification process are published which allow for computing the desired tissue activities. Alternatively, enzyme tissue activities may be computed if the specific activity (per gram of enzyme) and enzyme concentration in the tissue of interest are known. Frequently, measurements of enzyme kinetic data have been performed under conditions different from the ones in living tissues. Therefore, suitable corrections of measured kinetic data need to be applied, in particular for deviations in pH and temperature. After a model has been set up, it needs to be validated. To that end, physiological parameters like local substrate concentrations, partial pressures, or turnover rates must be identified the numerical values of which are known from the literature or can be easily measured on the one hand and which can also be predicted by the model on the other. Validation is performed by comparing the behaviour of measured and predicted values for a number of locations, physiological conditions or the like. In order to identify suitable test cases for the model, data and accession tools of such data repositories need to be structured in an appropriate fashion. In particular, the fields describing identified differential proteins need to be rendered searchable and a search mechanism has to be implemented that allows finding all experimental conditions under which a given protein of interest is up- or downregulated, changes its activity, undergoes certain modifications, etc.
In a recent example, in age-related comparisons of a number of model organisms, several enzymes of the dopamine metabolism turned out to be differential (19). Therefore, as a first model case, degradation of dopamine was selected. Dopamine formation and
48
Schrattenholz, Groebe, and Soskic
decomposition is part of the tyrosine metabolism. The dopaminerelated details of the pertinent pathway map by KEGG (110) (http://www.genome.jp/dbget-bin/get_pathway?org_ name=hsa&mapno=00350) is shown in Fig. 2. Dopamine is produced from l-dopa by the enzyme dopa decarboxylase (4.1.1.28). The subsequent dopamine metabolism is fairly complicated and involves the formation of the hormones epinephrine and norepinephrine. One of the enzymes found to be affected by ageing was a homologue of mammalian catecholamine O-methyltransferase (COMT). Also involved is monoamine oxidase (MAO) which plays an important role in the pharmacological modification of dopamine degradation. As becomes apparent from Fig. 2, dopamine metabolism is only linked to the rest of tyrosine metabolism via the enzyme dopa decarboxylase (4.1.1.28), so all reactions downstream of dopa decarboxylase may be viewed as a (largely) independent sub-network. Dopamine metabolism was modelled using the “Systems Biology Workbench” and JDesigner (113). The SBW framework allows for graphically setting up networks of chemical species
Fig. 2. Detail of KEGG pathway map of tyrosine metabolism (http://www.genome.jp/dbget-bin/get_pathway?org_ name=hsa&mapno=00350). Only reactions relating to dopamine formation and metabolism are shown. Dopamine is produced from l-dopa by the enzyme dopa decarboxylase (4.1.1.28). The subsequent dopamine metabolism is fairly complicated and involves the formation of the hormones epinephrine and norepinephrine. A homologue of catecholamine O-methyltransferase (COMT, 2.1.1.6) has been found to be upregulated in ageing. Also involved is monoamine oxidase (MAO, 1.4.3.4) which plays an important role in the pharmacological modification of dopamine degradation. Note that COMT as well as MAO participate in no fewer than six distinct reactions in the degradation of dopamine, epinephrine and norepinephrine.
Systems Biology Approaches and Tools for Analysis of Interactomes
49
Fig. 3. Graphical representation of the mathematical model of dopamine metabolism as displayed in JDesigner/Systems Biology Workbench. Each node in the network denotes a chemical species which takes part in the reactions. Each arrow corresponds to an enzyme, the reaction kinetics of which can be chosen from a variety of different kinetic laws with deliberately definable kinetic constants.
which take part in the reactions (Fig. 3). Each arrow represents an enzyme, the reaction kinetics of which can be chosen from a variety of different kinetic laws (or completely re-defined by the user) with kinetic constants determinable at one’s own discretion. In conclusion, proteins change compartments or active states due to posttranslational modifications like methylation, phosphorylation, glycosylation and non-enzymatic oxidation. Moreover, proteolytic processing occurs, inactivating or mobilizing effector domains. The kinetics of these events include and integrate fast (activity-dependent) and slow (developmental) time frames. Even in the apparent absence of posttranslational modifications dramatic localisational effects in completely different cells or compartments regulate cellular biology on the level of the organism. In order to reflect these various kinetic changes in protein function more closely, a number of prerequisites need to be
50
Schrattenholz, Groebe, and Soskic
met in future approaches to mathematical modelling of metabolic networks: ●
●
●
●
Necessity of kinetic control of biological material. We need time-resolved recordings of kinetically evolving synchronous, statistically significant differences in the proteomes of the compared samples. Necessity of precise knowledge concerning fractionation strategies of the biological material. Necessity of high resolution, large dynamic range, precise quantification, and meaningful statistics (84). Necessity of novel strategies in data analysis and interpretation. We need to identify signatures of redundant protein isoforms which are to be viewed as effect “spaces” or “landscapes.” On top of traditional bioinformatic analysis, the stochastic and non-linear properties of these effect spaces need mathematical attention.
These new approaches are expected to have most important consequences in drug discovery where we may face a paradigm shift towards low affinity/low selectivity drugs or even multidimensional (= dirty) drugs or drug combinations. Over and beyond the role of the proteome in pharmacological research, many novel treatments promoted under the heading “Biologics” or the like are anticipated to be proteins themselves. As a consequence, future system biologic strategies will need to cope with the following problems: ●
●
●
How to screen for multi-faceted kinetic proteome data? How to extract higher level explicative information by applying advanced mathematical models to the measured data? How to interpret the greatly complex, multidimensional modelling results?
5. Chemical Proteomics An important recent development in addressing problems associated with limited experimental windows due to poor resolution or dynamic range of separation and detection methodologies is the application of affinity-based approaches. Specific antibodies have been used to pull down targets and associated proteins. However, antibodies usually are selected against non-modified peptides and the complexity resulting from posttranslational modifications cannot be addressed. Moreover, cross-reactivity of antibodies and the fact that they are proteins themselves is causing problems in mass spectrometry-based identification procedures.
Systems Biology Approaches and Tools for Analysis of Interactomes
51
Fig. 4. General principle of affinity-based chemical proteomics. Protein labelling is three-step process. In the first step, affinity probe that consists of: (1) a reactive group, typically an electrophilic or photoreactive group, for covalent labelling of protein targets; (2) a binding group, which directs the reactive element towards specific classes of proteins; and (3) azide group that facilitates application bioorthogonal “click chemistry,” react with protein. In the second step, analytical probe that consists of: (1) fluorophore or (and) biotin, for detection and/or enrichment of probe-labelled proteins; and (2) acetylene group as counterpart to azide group for “click chemistry,” react with tagged protein. And, finally, in the third step labelled proteome can be selectively detected and enriched.
“Chemical proteomics,” is employing active site-directed chemical affinity probes to explore “interacting proteomes” for drugs/ligands as a special embodiment of affinity fractionation (30, 116) (Fig. 4). These affinity probes can either be generated by synthesizing ligands which are covalently bound to solid phases to create affinity resins for protein-binding partners or as soluble compounds consisting of at least two general elements: (1) a reactive group for binding and covalently modifying the binding site of ligand, and (2) a reporter tag for the detection, enrichment, and identification of probe-labelled proteins. Typical reporter tags include fluorophores and/or biotin for in-gel detection and avidinbased enrichment of probe-labelled enzymes or receptors, respectively. To date, this type of technology has been used successfully to target all major classes of proteases, kinases, phosphatases, glycosidases, GSTs, oxidoreductases, phosphatidyl-inositol 3-kinase signalling and histone deacetylases as well as for post-translationally modified proteins (24, 117–119). Bulky reporter tags, such as fluorophores or biotin, can be replaced with much smaller azide or alkyne groups which enable profiling of enzyme activities in living cells and animals (30, 120). In these experiments, post-labelling
52
Schrattenholz, Groebe, and Soskic
ex vivo conjugation to reporter tags is accomplished by a bioorthogonal reaction like the azide–alkyne cycloaddition or the Staudinger ligation (121). A recent example is the exploration of an additional mode of action of 4-azasteroids, like Dutasteride, which have been shown to interfere with the mitochondrial permeability transition pore (30).
6. Conclusion Drug development is currently moving forward from the nucleic acid-based high-throughput screening strategies of the past decades and examining novel concepts like systems biology beyond the “one disease-one-target and drug” thinking. On the contrary, latest genomic screening of ever increasing human populations rather indicate that many if not most pathological phenotypes may be converging from a multitude of subtle genomic changes. In a context of the resulting discussion about “personalized medicine” (122), it is increasingly acknowledged that successful compounds do not exert their effects through a single target, but instead have multiple targets and also rather have their effects on systemic phenotypes or key nodal points of integrative molecular/biochemical networks, downstream of genes and/or SNP’s. There is hope that these additional modes of action of compounds with lower selectivities and affinities might have a better chance of positive effects in complex equilibriums of whole cellular networks (118). However, the question of rational design and screening of such multi-target drugs remains open (119). On a first level, the challenges concerning robust tools for screening and characterization of corresponding protein biomarkers are a hurdle. Once statistically and biologically validated biomarker signatures for compounds across a relevant set of biological conditions have been firmly established, it is not yet decided how to employ this type of complex information for screening. The integration of experimental and modelling approaches is the major challenge of the emerging field of systems biology. The conceptual frameworks for generating quantitative data relevant on a systems level, the appropriate statistical and mathematical treatment of such data sets and subsequent structural and pathways modelling still do not exist. The mere quantitative reconstruction of the dynamics of cellular molecular changes is still lacking generally accepted experimental approaches and strategies. The degrees of freedom of metastable biological systems and the information codes underlying these disequilibrium states are maintained by energetically extremely costly regulation (ATP, NADH).
Systems Biology Approaches and Tools for Analysis of Interactomes
53
These tools could help to identify a suitable set of parallel targets and multi-target substances for specific conditions. The integrative character of systems biology with a strong focus on the proteomic level is already having some impact on the drug discovery process. Future screening procedures will probably be not so much high throughput but will include opportunistic wet laboratory optimization by combining in vitro models with chemical proteomics and affinity fractionation (120, 123). Drugs will be characterized by selected endpoints of more complex (76) (Fig. 3) dynamic signatures of “interacting proteomes” (84). Many of them will be biologics, due to the modular structural flexibility of biomolecules. References 1. Boone C, Bussey H, Andrews BJ (2007) Exploring genetic interactions and networks with yeast. Nat Rev Genet 8:437–449 2. Davierwala AP, Haynes J, Li Z, Brost RL, Robinson MD, Yu L, Mnaimneh S, Ding H, Zhu H, Chen Y, Cheng X, Brown GW, Boone C, Andrews BJ, Hughes TR (2005) The synthetic genetic interaction spectrum of essential genes. Nat Genet 37:1147–1152 3. Tong AH, Lesage G, Bader GD, Ding H, Xu H, Xin X, Young J, Berriz GF, Brost RL, Chang M, Chen Y, Cheng X, Chua G, Friesen H, Goldberg DS, Haynes J, Humphries C, He G, Hussein S, Ke L, Krogan N, Li Z, Levinson JN, Lu H, Menard P, Munyana C, Parsons AB, Ryan O, Tonikian R, Roberts T, Sdicu AM, Shapiro J, Sheikh B, Suter B, Wong SL, Zhang LV, Zhu H, Burd CG, Munro S, Sander C, Rine J, Greenblatt J, Peter M, Bretscher A, Bell G, Roth FP, Brown GW, Andrews B, Bussey H, Boone C (2004) Global mapping of the yeast genetic interaction network. Science 303:808–813 4. Hughes TR, Robinson MD, Mitsakakis N, Johnston M (2004) The promise of functional genomics: completing the encyclopedia of a cell. Curr Opin Microbiol 7:546–554 5. Dolinski K, Botstein D (2005) Changing perspectives in yeast research nearly a decade after the genome sequence. Genome Res 15:1611–1619 6. Lehner B, Crombie C, Tischler J, Fortunato A, Fraser AG (2006) Systematic mapping of genetic interactions in Caenorhabditis elegans identifies common modifiers of diverse signaling pathways. Nat Genet 38:896–903 7. Schlabach MR, Luo J, Solimini NL, Hu G, Xu Q, Li MZ, Zhao Z, Smogorzewska A, Sowa ME, Ang XL, Westbrook TF, Liang AC, Chang K, Hackett JA, Harper JW,
Hannon GJ, Elledge SJ (2008) Cancer proliferation gene discovery through functional genomics. Science 319:620–624 8. Silva JM, Marran K, Parker JS, Silva J, Golding M, Schlabach MR, Elledge SJ, Hannon GJ, Chang K (2008) Profiling essential genes in human mammary cells by multiplex RNAi screening. Science 319:617–620 9. Berns K, Horlings HM, Hennessy BT, Madiredjo M, Hijmans EM, Beelen K, Linn SC, Gonzalez-Angulo AM, Stemke-Hale K, Hauptmann M, Beijersbergen RL, Mills GB, van de V, Bernards R (2007) A functional genetic approach identifies the PI3K pathway as a major determinant of trastuzumab resistance in breast cancer. Cancer Cell 12:395–402 10. Cordell HJ (2002) Epistasis: what it means, what it doesn’t mean, and statistical methods to detect it in humans. Hum Mol Genet 11:2463–2468 11. Luo B, Cheung HW, Subramanian A, Sharifnia T, Okamoto M, Yang X, Hinkle G, Boehm JS, Beroukhim R, Weir BA, Mermel C, Barbie DA, Awad T, Zhou X, Nguyen T, Piqani B, Li C, Golub TR, Meyerson M, Hacohen N, Hahn WC, Lander ES, Sabatini DM, Root DE (2008) Highly parallel identification of essential genes in cancer cells. Proc Natl Acad Sci USA 105:20380–20385 12. Ito T, Chiba T, Ozawa R, Yoshida M, Hattori M, Sakaki Y (2001) A comprehensive twohybrid analysis to explore the yeast protein interactome. Proc Natl Acad Sci USA 98:4569–4574 13. Uetz P, Giot L, Cagney G, Mansfield TA, Judson RS, Knight JR, Lockshon D, Narayan V, Srinivasan M, Pochart P, Qureshi-Emili A, Li Y, Godwin B, Conover D, Kalbfleisch T, Vijayadamodar G, Yang M, Johnston M, Fields S, Rothberg JM (2000) A comprehensive
54
Schrattenholz, Groebe, and Soskic
analysis of protein-protein interactions in Saccharomyces cerevisiae. Nature 403:623–627 14. Ho Y, Gruhler A, Heilbut A, Bader GD, Moore L, Adams SL, Millar A, Taylor P, Bennett K, Boutilier K, Yang L, Wolting C, Donaldson I, Schandorff S, Shewnarane J, Vo M, Taggart J, Goudreault M, Muskat B, Alfarano C, Dewar D, Lin Z, Michalickova K, Willems AR, Sassi H, Nielsen PA, Rasmussen KJ, Andersen JR, Johansen LE, Hansen LH, Jespersen H, Podtelejnikov A, Nielsen E, Crawford J, Poulsen V, Sorensen BD, Matthiesen J, Hendrickson RC, Gleeson F, Pawson T, Moran MF, Durocher D, Mann M, Hogue CW, Figeys D, Tyers M (2002) Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 415:180–183 15. Krogan NJ, Cagney G, Yu H, Zhong G, Guo X, Ignatchenko A, Li J, Pu S, Datta N, Tikuisis AP, Punna T, Peregrin-Alvarez JM, Shales M, Zhang X, Davey M, Robinson MD, Paccanaro A, Bray JE, Sheung A, Beattie B, Richards DP, Canadien V, Lalev A, Mena F, Wong P, Starostine A, Canete MM, Vlasblom J, Wu S, Orsi C, Collins SR, Chandran S, Haw R, Rilstone JJ, Gandi K, Thompson NJ, Musso G, St OP, Ghanny S, Lam MH, Butland G, Altaf-Ul AM, Kanaya S, Shilatifard A, O’Shea E, Weissman JS, Ingles CJ, Hughes TR, Parkinson J, Gerstein M, Wodak SJ, Emili A, Greenblatt JF (2006) Global landscape of protein complexes in the yeast Saccharomyces cerevisiae. Nature 440:637–643 16. Gavin AC, Bosche M, Krause R, Grandi P, Marzioch M, Bauer A, Schultz J, Rick JM, Michon AM, Cruciat CM, Remor M, Hofert C, Schelder M, Brajenovic M, Ruffner H, Merino A, Klein K, Hudak M, Dickson D, Rudi T, Gnau V, Bauch A, Bastuck S, Huhse B, Leutwein C, Heurtier MA, Copley RR, Edelmann A, Querfurth E, Rybin V, Drewes G, Raida M, Bouwmeester T, Bork P, Seraphin B, Kuster B, Neubauer G, SupertiFurga G (2002) Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 415:141–147 17. Winzeler EA, Shoemaker DD, Astromoff A, Liang H, Anderson K, Andre B, Bangham R, Benito R, Boeke JD, Bussey H, Chu AM, Connelly C, Davis K, Dietrich F, Dow SW, El BM, Foury F, Friend SH, Gentalen E, Giaever G, Hegemann JH, Jones T, Laub M, Liao H, Liebundguth N, Lockhart DJ, Lucau-Danila A, Lussier M, M’Rabet N, Menard P, Mittmann M, Pai C, Rebischung C, Revuelta JL, Riles L, Roberts CJ, RossMacDonald P, Scherens B, Snyder M, Sookhai-Mahadeo S, Storms RK, Veronneau
S, Voet M, Volckaert G, Ward TR, Wysocki R, Yen GS, Yu K, Zimmermann K, Philippsen P, Johnston M, Davis RW (1999) Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285:901–906 18. Maxwell CA, Moreno V, Sole X, Gomez L, Hernandez P, Urruticoechea A, Pujana MA (2008) Genetic interactions: the missing links for a better understanding of cancer susceptibility, progression and treatment. Mol Cancer 7:4 19. Groebe K, Krause F, Kunstmann B, Unterluggauer H, Reifschneider NH, Scheckhuber CQ, Sastri C, Stegmann W, Wozny W, Schwall GP, Poznanovic S, Dencher NA, Jansen-Durr P, Osiewacz HD, Schrattenholz A (2007) Differential proteomic profiling of mitochondria from Podospora anserina, rat and human reveals distinct patterns of age-related oxidative changes. Exp Gerontol 42:887–898 20. Parsons AB, Brost RL, Ding H, Li Z, Zhang C, Sheikh B, Brown GW, Kane PM, Hughes TR, Boone C (2004) Integration of chemical-genetic and genetic interaction data links bioactive compounds to cellular target pathways. Nat Biotechnol 22:62–69 21. Sharom JR, Bellows DS, Tyers M (2004) From large networks to small molecules. Curr Opin Chem Biol 8:81–90 22. Burkard ME, Randall CL, Larochelle S, Zhang C, Shokat KM, Fisher RP, Jallepalli PV (2007) Chemical genetics reveals the requirement for Polo-like kinase 1 activity in positioning RhoA and triggering cytokinesis in human cells. Proc Natl Acad Sci USA 104:4383–4388 23. Larochelle S, Merrick KA, Terret ME, Wohlbold L, Barboza NM, Zhang C, Shokat KM, Jallepalli PV, Fisher RP (2007) Requirements for Cdk7 in the assembly of Cdk1/cyclin B and activation of Cdk2 revealed by chemical genetics in human cells. Mol Cell 25:839–850 24. Schrattenholz A, Soskic V (2008) What does systems biology mean for drug development? Curr Top Med Chem 15:1520–1528 25. Ashburn TT, Thor KB (2004) Drug repositioning: identifying and developing new uses for existing drugs. Nat Rev Drug Discov 3:673–683 26. Youdim MB (2007) Magic bullets or novel multimodal drugs with various CNS targets for Parkinson’s disease? Nat Rev Drug Discov 6:499–500 27. Natt F (2007) siRNAs in drug discovery: target validation and beyond. Curr Opin Mol Ther 9:242–247
Systems Biology Approaches and Tools for Analysis of Interactomes 28. Burbaum J, Tobal GM (2002) Proteomics in drug discovery. Curr Opin Chem Biol 6:427–433 29. Keskin O, Gursoy A, Ma B, Nussinov R (2007) Towards drugs targeting multiple proteins in a systems biology approach. Curr Top Med Chem 7:943–951 30. Soskic V, Klemm M, Proikas-Cezanne T, Schwall GP, Poznanovic S, Stegmann W, Groebe K, Zengerling H, Schoepf R, Burnet M, Schrattenholz A (2007) A connection between the mitochondrial permeability transition pore, autophagy and cerebral amyloidogenesis. J Proteome Res 7:2262–2269 31. Drews J, Ryser S (1997) The role of innovation in drug development. Nat Biotechnol 15:1318–1319 32. Imming P, Sinning C, Meyer A (2006) Drugs, their targets and the nature and number of drug targets. Nat Rev Drug Discov 5:821–834 33. Schrattenholz A, Soskic V (2006) NMDA receptors are not alone: dynamic regulation of NMDA receptor structure and function by neuregulins and transient cholesterol-rich membrane domains leads to disease-specific nuances of glutamate-signalling. Curr Top Med Chem 6:663–686 34. Apic G, Ignjatovic T, Boyer S, Russell RB (2005) Illuminating drug discovery with biological pathways. FEBS Lett 579:1872–1877 35. Huang S, Ingber DE (2006) A non-genetic basis for cancer progression and metastasis: self-organizing attractors in cell regulatory networks. Breast Dis 26:27–54 36. Tsuda I, Fujii H (2007) Chaos reality in the brain. J Integr Neurosci 6:309–326 37. Schoner G, Scholz JP (2007) Analyzing variance in multi-degree-of-freedom movements: uncovering structure versus extracting correlations. Mot Control 11:259–275 38. Southern J, Pitt-Francis J, Whiteley J, Stokeley D, Kobashi H, Nobes R, Kadooka Y, Gavaghan D (2008) Multi-scale computational modelling in biology and physiology. Prog Biophys Mol Biol 96:60–89 39. Price ND, Shmulevich I (2007) Biochemical and statistical network models for systems biology. Curr Opin Biotechnol 18:365–370 40. Konopka AK (2007) Surrogacy theory and models of convoluted organic systems. Proteomics 7:846–856 41. Huang S, Wikswo J (2006) Dimensions of systems biology. Rev Physiol Biochem Pharmacol 157:81–104 42. Boelsterli UA, Lim PL (2007) Mitochondrial abnormalities – a link to idiosyncratic drug
55
hepatotoxicity? Toxicol Appl Pharmacol 220:92–107 43. Dykens JA, Marroquin LD, Will Y (2007) Strategies to reduce late-stage drug attrition due to mitochondrial toxicity. Expert Rev Mol Diagn 7:161–175 44. Lecellier G, Brenner C (2007) Genomic and proteomic screening of apoptosis mitochondrial regulators for drug target discovery. Curr Med Chem 14:875–881 45. Roses AD, Saunders AM, Huang Y, Strum J, Weisgraber KH, Mahley RW (2007) Complex disease-associated pharmacogenetics: drug efficacy, drug safety, and confirmation of a pathogenetic hypothesis (Alzheimer’s disease). Pharmacogenomics J 7:10–28 46. Jonker DM, Visser SA, van der Graaf PH, Voskuyl RA, Danhof M (2005) Towards a mechanism-based analysis of pharmacodynamic drug-drug interactions in vivo. Pharmacol Ther 106:1–18 47. Harman D (1956) Aging: a theory based on free radical and radiation chemistry. J Gerontol 11:298–300 48. Ward M (2007) Biomarkers for Alzheimer’s disease. Expert Rev Mol Diagn 7:635–646 49. Filley CM, Rollins YD, Anderson CA, Arciniegas DB, Howard KL, Murrell JR, Boyer PJ, Kleinschmidt-DeMasters BK, Ghetti B (2007) The genetics of very early onset Alzheimer disease. Cogn Behav Neurol 20:149–156 50. Sonnen JA, Keene CD, Montine KS, Li G, Peskind ER, Zhang J, Montine TJ (2007) Biomarkers for Alzheimer’s disease. Expert Rev Neurother 7:1021–1028 51. Ertekin-Taner N (2007) Genetics of Alzheimer’s disease: a centennial review. Neurol Clin 25:611–667, v 52. Chai CK (2007) The genetics of Alzheimer’s disease. Am J Alzheimers Dis Other Demen 22:37–41 53. Mancuso M, Coppede F, Murri L, Siciliano G (2007) Mitochondrial cascade hypothesis of Alzheimer’s disease: myth or reality? Antioxid Redox Signal 9:1631–1646 54. Reid PC, Urano Y, Kodama T, Hamakubo T (2007) Alzheimer’s disease: cholesterol, membrane rafts, isoprenoids and statins. J Cell Mol Med 11:383–392 55. Argyle DJ, Blacking T (2008) From viruses to cancer stem cells: dissecting the pathways to malignancy. Vet J 177:311–323 56. Sales KM, Winslet MC, Seifalian AM (2007) Stem cells and cancer: an overview. Stem Cell Rev 3:249–255
56
Schrattenholz, Groebe, and Soskic
57. Alison MR, Murphy G, Leedham S (2008) Stem cells and cancer: a deadly mix. Cell Tissue Res 331:109–124 58. Erenpreisa J, Cragg MS (2007) Cancer: a matter of life cycle? Cell Biol Int 31:1507–1510 59. Sjoblom T, Jones S, Wood LD, Parsons DW, Lin J, Barber TD, Mandelker D, Leary RJ, Ptak J, Silliman N, Szabo S, Buckhaults P, Farrell C, Meeh P, Markowitz SD, Willis J, Dawson D, Willson JK, Gazdar AF, Hartigan J, Wu L, Liu C, Parmigiani G, Park BH, Bachman KE, Papadopoulos N, Vogelstein B, Kinzler KW, Velculescu VE (2006) The consensus coding sequences of human breast and colorectal cancers. Science 314:268–274 60. Venkatesan RN, Bielas JH, Loeb LA (2006) Generation of mutator mutants during carcinogenesis. DNA Repair (Amst) 5:294–302 61. Araujo RP, Liotta LA, Petricoin EF (2007) Proteins, drug targets and the mechanisms they control: the simple truth about complex networks. Nat Rev Drug Discov 6:871–880 62. Lehar J, Zimmermann GR, Krueger AS, Molnar RA, Ledell JT, Heilbut AM, Short GF III, Giusti LC, Nolan GP, Magid OA, Lee MS, Borisy AA, Stockwell BR, Keith CT (2007) Chemical combination effects predict connectivity in biological systems. Mol Syst Biol 3:80 63. Avis I, Martinez A, Tauler J, Zudaire E, Mayburd A, Abu-Ghazaleh R, Ondrey F, Mulshine JL (2005) Inhibitors of the arachidonic acid pathway and peroxisome proliferator-activated receptor ligands have superadditive effects on lung cancer growth inhibition. Cancer Res 65:4181–4190 64. Liby KT, Yore MM, Sporn MB (2007) Triterpenoids and rexinoids as multifunctional agents for the prevention and treatment of cancer. Nat Rev Cancer 7:357–369 65. Grosser T, Fries S, FitzGerald GA (2006) Biological basis for the cardiovascular consequences of COX-2 inhibition: therapeutic challenges and opportunities. J Clin Invest 116:4–15 66. Frantz S (2005) Drug discovery: playing dirty. Nature 437:942–943 67. Millan MJ (2006) Multi-target strategies for the improved treatment of depressive states: conceptual foundations and neuronal substrates, drug discovery and therapeutic application. Pharmacol Ther 110:135–370 68. Farah A (2005) Atypicality of atypical antipsychotics. Prim Care Companion J Clin Psychiatry 7:268–274 69. Roth BL, Sheffler DJ, Kroeze WK (2004) Magic shotguns versus magic bullets: selectively non-selective drugs for mood disorders
and schizophrenia. Nat Rev Drug Discov 3:353–359 70. Stahl SM (1998) What makes an antipsychotic atypical? J Clin Psychiatry 59:403–404 71. Ahmad N, Feyes DK, Nieminen AL, Agarwal R, Mukhtar H (1997) Green tea constituent epigallocatechin-3-gallate and induction of apoptosis and cell cycle arrest in human carcinoma cells. J Natl Cancer Inst 89: 1881–1886 72. Khan N, Afaq F, Saleem M, Ahmad N, Mukhtar H (2006) Targeting multiple signaling pathways by green tea polyphenol (−)-epigallocatechin-3-gallate. Cancer Res 66:2500–2505 73. Khan N, Afaq F, Mukhtar H (2007) Apoptosis by dietary factors: the suicide solution for delaying cancer growth. Carcinogenesis 28:233–239 74. Hopkins AL (2008) Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol 4:682–690 75. Dessalew N, Workalemahu M (2008) On the paradigm shift towards multitarget selective drug design. Curr Comput Aided Drug Des 4:76–90 76. Korcsmaros T, Szalay MS, Bode C, Kovacs IA, Csermely P (2007) How to design multitarget drugs: target search options in cellular networks. Expert Opin Drug Discov 2:1–10 77. Barabasi AL, Albert R (1999) Emergence of scaling in random networks. Science 286:509–512 78. Palla G, Derenyi I, Farkas I, Vicsek T (2005) Uncovering the overlapping community structure of complex networks in nature and society. Nature 435:814–818 79. Nacher JC, Schwartz JM (2008) A global view of drug-therapy interactions. BMC Pharmacol 8:5 80. Csermely P, Agoston V, Pongor S (2005) The efficiency of multi-target drugs: the network approach might help drug design. Trends Pharmacol Sci 26:178–182 81. Strong M, Eisenberg D (2007) The protein network as a tool for finding novel drug targets. Prog Drug Res 64:191, 193–215 82. Morphy R, Rankovic Z (2007) Fragments, network biology and designing multiple ligands. Drug Discov Today 12:156–160 83. Hyduke DR, Jarboe LR, Tran LM, Chou KJ, Liao JC (2007) Integrated network analysis identifies nitric oxide response networks and dihydroxyacid dehydratase as a crucial target in Escherichia coli. Proc Natl Acad Sci USA 104:8484–8489
Systems Biology Approaches and Tools for Analysis of Interactomes 84. Schrattenholz A, Groebe K (2007) What does it need to be a biomarker? Relationships between resolution, differential quantification and statistical validation of protein surrogate biomarkers. Electrophoresis 28:1970–1979 85. Kovacs IA, Szalay MS, Csermely P (2005) Water and molecular chaperones act as weak links of protein folding networks: energy landscape and punctuated equilibrium changes point towards a game theory of proteins. FEBS Lett 579:2254–2260 86. Shen-Orr SS, Milo R, Mangan S, Alon U (2002) Network motifs in the transcriptional regulation network of Escherichia coli. Nat Genet 31:64–68 87. Jiang R, Tu Z, Chen T, Sun F (2006) Network motif identification in stochastic networks. Proc Natl Acad Sci USA 103:9404–9409 88. Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D, Alon U (2002) Network motifs: simple building blocks of complex networks. Science 298:824–827 89. Milo R, Itzkovitz S, Kashtan N, Levitt R, Shen-Orr S, Ayzenshtat I, Sheffer M, Alon U (2004) Superfamilies of evolved and designed networks. Science 303:1538–1542 90. Agoston V, Csermely P, Pongor S (2005) Multiple weak hits confuse complex systems: a transcriptional regulatory network as an example. Phys Rev E Stat Nonlin Soft Matter Phys 71:051909 91. Sivachenko AY, Yuryev A (2007) Pathway analysis software as a tool for drug target selection, prioritization and validation of drug mechanism. Expert Opin Ther Targets 11:411–421 92. Herwig R, Lehrach H (2006) Expression profiling of drug response – from genes to pathways. Dialogues Clin Neurosci 8:283–293 93. Elrick MM, Walgren JL, Mitchell MD, Thompson DC (2006) Proteomics: recent applications and new technologies. Basic Clin Pharmacol Toxicol 98:432–441 94. Schrattenholz A, Klemm M (2006) How human embryonic stem cell research can impact in vitro drug screening technologies of the future. In: Marx U, Sandig V (eds) Drug testing in vitro: breakthroughs and trends in cell culture technology. Wiley/ VCH, New York, pp 205–228 95. Schrattenholz A, Klemm M (2007) Neuronal cell culture from human embryonic stem cells as in vitro model for neuroprotection. ALTEX 24:9–15 96. Malmstrom J, Lee H, Aebersold R (2007) Advances in proteomic workflows for systems biology. Curr Opin Biotechnol 18:378–384
57
97. Hunzinger C, Schrattenholz A, Poznanovic S, Schwall GP, Stegmann W (2006) Comparison of different separation technologies for proteome analyses: isoform resolution as a prerequisite for the definition of protein biomarkers on the level of posttranslational modifications. J Chromatogr A 1123:170–181 98. Ho E, Webber R, Wilkins MR (2007) Interactive three-dimensional visualization and contextual analysis of protein interaction networks. J Proteome Res 7:104–112 99. Grillari J, Katinger H, Voglauer R (2006) Aging and the ubiquitinome: traditional and non-traditional functions of ubiquitin in aging cells and tissues. Exp Gerontol 41:1067–1079 100. Tanaka K (2009) The proteasome: overview of structure and functions. Proc Jpn Acad B Phys Biol Sci 85:12–36 101. Ulrich HD (2009) The SUMO system: an overview. Methods Mol Biol 497:3–16 102. Schrattenholz A, Šoškic´ V, Groebe K (2010) Synchronisation of posttranslational modifications during ageing: time is a crucial biological dimension. Ann NY Acad Sci 1197:118–128 103. Brazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman P, Stoeckert C, Aach J, Ansorge W, Ball CA, Causton HC, Gaasterland T, Glenisson P, Holstege FC, Kim IF, Markowitz V, Matese JC, Parkinson H, Robinson A, Sarkans U, Schulze-Kremer S, Stewart J, Taylor R, Vilo J, Vingron M (2001) Minimum information about a microarray experiment (MIAME)-toward standards for microarray data. Nat Genet 29:365–371 104. Taylor CF, Field D, Sansone SA, Aerts J, Apweiler R, Ashburner M, Ball CA, Binz PA, Bogue M, Booth T, Brazma A, Brinkman RR, Michael CA, Deutsch EW, Fiehn O, Fostel J, Ghazal P, Gibson F, Gray T, Grimes G, Hancock JM, Hardy NW, Hermjakob H, Julian RK Jr, Kane M, Kettner C, Kinsinger C, Kolker E, Kuiper M, Le NN, LeebensMack J, Lewis SE, Lord P, Mallon AM, Marthandan N, Masuya H, McNally R, Mehrle A, Morrison N, Orchard S, Quackenbush J, Reecy JM, Robertson DG, Rocca-Serra P, Rodriguez H, Rosenfelder H, Santoyo-Lopez J, Scheuermann RH, Schober D, Smith B, Snape J, Stoeckert CJ Jr, Tipton K, Sterk P, Untergasser A, Vandesompele J, Wiemann S (2008) Promoting coherent minimum reporting guidelines for biological and biomedical investigations: the MIBBI project. Nat Biotechnol 26:889–896 105. Taylor CF, Paton NW, Lilley KS, Binz PA, Julian RK Jr, Jones AR, Zhu W, Apweiler R, Aebersold R, Deutsch EW, Dunn MJ, Heck AJ,
58
Schrattenholz, Groebe, and Soskic
Leitner A, Macht M, Mann M, Martens L, Neubert TA, Patterson SD, Ping P, Seymour SL, Souda P, Tsugita A, Vandekerckhove J, Vondriska TM, Whitelegge JP, Wilkins MR, Xenarios I, Yates JR III, Hermjakob H (2007) The minimum information about a proteomics experiment (MIAPE). Nat Biotechnol 25:887–893 106. Jones AR, Miller M, Aebersold R, Apweiler R, Ball CA, Brazma A, Degreef J, Hardy N, Hermjakob H, Hubbard SJ, Hussey P, Igra M, Jenkins H, Julian RK Jr, Laursen K, Oliver SG, Paton NW, Sansone SA, Sarkans U, Stoeckert CJ Jr, Taylor CF, Whetzel PL, White JA, Spellman P, Pizarro A (2007) The Functional Genomics Experiment model (FuGE): an extensible framework for standards in functional genomics. Nat Biotechnol 25:1127–1133 107. Kaiser J (2002) Proteomics. Public-private group maps out initiatives. Science 296:827 108. Orchard S, Hermjakob H, Apweiler R (2003) The proteomics standards initiative. Proteomics 3:1374–1376 109. Jones P, Cote RG, Martens L, Quinn AF, Taylor CF, Derache W, Hermjakob H, Apweiler R (2006) PRIDE: a public repository of protein and peptide identifications for the proteomics community. Nucleic Acids Res 34:D659–D663 110. Kanehisa M, Goto S, Kawashima S, Okuno Y, Hattori M (2004) The KEGG resource for deciphering the genome. Nucleic Acids Res 32:D277–D280 111. Junker BH, Klukas C, Schreiber F (2006) VANTED: a system for advanced data analysis and visualization in the context of biological networks. BMC Bioinform 7:109 112. Schomburg I, Chang A, Schomburg D (2002) BRENDA, enzyme data and metabolic information. Nucleic Acids Res 30:47–49 113. Sauro HM, Hucka M, Finney A, Wellock C, Bolouri H, Doyle J, Kitano H (2003) Next generation simulation tools: the Systems Biology Workbench and BioSPICE integration. OMICS 7:355–372 114. Hucka M, Finney A, Sauro HM, Bolouri H, Doyle JC, Kitano H, Arkin AP, Bornstein BJ, Bray D, Cornish-Bowden A, Cuellar AA,
Dronov S, Gilles ED, Ginkel M, Gor V, Goryanin II, Hedley WJ, Hodgman TC, Hofmeyr JH, Hunter PJ, Juty NS, Kasberger JL, Kremling A, Kummer U, Le NN, Loew LM, Lucio D, Mendes P, Minch E, Mjolsness ED, Nakayama Y, Nelson MR, Nielsen PF, Sakurada T, Schaff JC, Shapiro BE, Shimizu TS, Spence HD, Stelling J, Takahashi K, Tomita M, Wagner J, Wang J (2003) The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models. Bioinformatics 19:524–531 115. Borger S, Liebermeister W, Klipp E (2006) Prediction of enzyme kinetic parameters based on statistical learning. Genome Inform 17:80–87 116. Simon GM, Cravatt BF (2008) Challenges for the ‘chemical-systems’ biologist. Nat Chem Biol 4:639–642 117. Salisbury CM, Cravatt BF (2008) Optimization of activity-based probes for proteomic profiling of histone deacetylase complexes. J Am Chem Soc 130:2184–2194 118. Rexach JE, Clark PM, Hsieh-Wilson LC (2008) Chemical approaches to understanding O-GlcNAc glycosylation in the brain. Nat Chem Biol 4:97–106 119. Martin BR, Cravatt BF (2009) Large-scale profiling of protein palmitoylation in mammalian cells. Nat Methods 6:135–138 120. Codelli JA, Baskin JM, Agard NJ, Bertozzi CR (2008) Second-generation difluorinated cyclooctynes for copper-free click chemistry. J Am Chem Soc 130:11486–11493 121. Ovaa H, van Swieten PF, Kessler BM, Leeuwenburgh MA, Fiebiger E, van den Nieuwendijk AM, Galardy PJ, van der Marel GA, Ploegh HL, Overkleeft HS (2003) Chemistry in living cells: detection of active proteasomes by a two-step labeling strategy. Angew Chem Int Ed Engl 42:3626–3629 122. Jorgensen JT (2009) New era of personalized medicine: a 10-year anniversary. Oncologist 14:557–558 123. Kurpiers T, Mootz HD (2009) Bioorthogonal ligation in the spotlight. Angew Chem Int Ed Engl 48:1729–1731
Chapter 3 Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics and RNAi Approach in Drosophila melanogaster Alessandro Cuomo and Tiziana Bonaldi Abstract Stable isotope labeling with amino acids in cell culture (SILAC) has become increasingly popular as a quantitative proteomics (qProteomics) method. In combination with high-resolution mass spectrometry (MS) and new efficient algorithms for the analysis of quantitative MS data, SILAC has proven to be a potent tool for the in-depth characterization of functional states. QProteomics extends transcriptomics analysis in providing comprehensive and unbiased protein expression profiles. In this chapter, we describe the use of SILAC procedure in combination with RNA interference (RNAi) to characterize loss-offunction phenotypes, an example to illustrate how qProteomics can address many of the systems-wide approaches previously restricted to the mRNA level. Furthermore, by explaining the adaptation of SILAC to a novel cellular model, the Drosophila melanogaster Schneider cells SL2, we aim to offer an example enabling the readers to apply the same strategy to any other cell culture, specific for their need. Key words: Quantitative proteomics, RNA interference, SILAC, Schneider cells, Mass spectrometry, Orbitrap, MaxQuant
1. Introduction Seeking differences between one functional state of a biological system versus another has been the source of many seminal discoveries. In order to increase the wealth of information achievable by such comparative analyses, comprehensive or system-wide screening approaches are needed. Differences between cellular states are reflected in changes in gene expression, at the level of both the message (mRNA) and the final product (protein). The first “genome-wide” method for expression analysis was offered by mRNA-based microarrays, which allow measuring thousands of transcripts in a single experiment. The major drawback of Qing Yan (ed.), Systems Biology in Drug Discovery and Development: Methods and Protocols, Methods in Molecular Biology, vol. 662, DOI 10.1007/978-1-60761-800-3_3, © Springer Science+Business Media, LLC 2010
59
60
Cuomo and Bonaldi
transcriptomics analysis is the lack of information at the level of proteins, the effectors of most biological functions. Protein levels, in fact, depend not only on the levels of the corresponding messages but also on a panel of translational controls and regulated degradation (1, 2). These factors may be as essential as mRNA synthesis, but they cannot be measured by microarrays (3–5). Thus, the information provided by proteomics offers a much closer description of the phenotype of a cell in a specific state (6–8). In spite of this enormous potential and of the remarkable progresses in mass spectrometry (MS), MS-based “shotgun” proteomics still faces great challenges: sensitivity and dynamic range are the most striking and have been thoroughly discussed (9–11). In case of comparative studies, not only identification, but also accurate and comprehensive quantitation of protein levels is of utmost importance. Whereas stable isotope labeling with amino acids in cell culture (SILAC) has proved extremely successful in several quantitative proteomics studies, the accurate and robust quantitation of proteins from MS-data still remains a very challenging task, limiting the applicability of quantitative strategies in system-wide screenings. Recently, the release of MaxQuant, a new efficient algorithm specifically developed for the analysis of high-resolution, quantitative MS data, represented a major breakthrough for the quantitation of proteins in SILAC-based experiments (12). Apart from achieving a considerable increase in peptide identification rates, MaxQuant allows the automatic quantification of several hundreds of thousands peptides per experiment and, consequently, the statistically robust quantification of several thousands proteins in complex proteomes. With such computational tool at hand, proteomics appears as the most informative instrument for “functional genomics” experiments, where microarrays have been so far the method of choice. One example of functional genomics studies, where systemwide screening produce a comprehensive picture of the cellular response to a perturbation, is represented by profiling gene expression upon ablation of a gene, or a gene product. Loss-offunction is a popular approach to gain functional insight in the large subset of annotated genes, whose physiological role is still elusive. Loss-of-function, traditionally achieved by classical forward genetics methods, has been facilitated by the development of RNA interference (RNAi) approach. Model systems such as Caenorhabditis elegans and Drosophila melanogaster are excellent organisms to perform RNAi, due to the possibility to deliver long double-stranded RNA (dsRNA). The easiness of RNA transfer, the high penetrance of target depletion, and the lower incidence of “off-targets” (13) rendered interference in those model organisms highly successful. Recently, a genome-wide library of double-stranded RNAs targeting every gene in the Drosophila
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
61
genome was published, paving the way to high-throughput genetic screens (14). In a systems biology perspective, comprehensive screenings describing the cellular response to RNAi are required: while microarrays have been largely employed in the past, quantitative proteomics is emerging as the elective choice for high-content phenotypization. Thus, a specific investment of qProteomics in model organisms such as worm and fly is desirable: with the possibility to use cell-based assays, SILAC is preferred. Recently, we published the first SILAC-labeling of D. melanogaster Schneider SL2 cultured cells, with the acquisition of quantitative transcriptome and proteome upon RNAi of a specific chromatin remodeling activity (ISWI) (15). We refer to complete reviews for the description of the basic principles of SILAC, for the recipe of standard SILAC media formulation (DMEM, RPMI) and for the conceptual explanation of MS-based quantitation (16, 17). We also strongly advise reading the article introducing the rationale and the algorithms for the quantitative analysis of SILAC data by MaxQuant (12). Instead in this chapter, we choose to focus on possible troubleshooting faced when adapting SILAC to novel model systems. For instance, we will discuss the reduction of cellular growth due to dialysis of complements in the medium, the limited incorporation and/or interconversion of heavy amino acids as potential sources of quantitation errors and the challenges in increasing the proteome coverage to gain statistical confidence in quantitation. Basing on our direct experience, we offer the “hints and tips” that facilitate the successful establishment of this quantitative strategy. Furthermore, we will describe the rationale for combining SILAC with RNAi in order to gain a systems biology view of lossof-function phenotypes, which is somewhat analogous to the microarray-based strategies, already extensively described. As such, we intend to present a “reference protocol” that enables the readers to adapt SILAC to their own cell culture models and for their own specific needs. The approach described is, in fact, generic and can be easily extended to other gene-silencing system (full knock-out, microRNA over expression, inhibitors) and/or to other model cell lines.
2. Materials 2.1. Standard Culture of SL2 Drosophila Embryonic Cells
1. Drosophila S2 Cells Schneider’s Drosophila Medium. 2. Fetal Bovine Serum (Heat Inactivated, insect cell culture tested).
62
Cuomo and Bonaldi
3. Penicillin 50 units/ml (Sigma). 4. Streptomycin 50 mg/ml (Sigma). 5. Bottles for growth of cells in suspension. 6. Cell culture dishes. 7. Incubator at 26°C, without CO2 supply. 2.2. SILAC Labeling of SL2 Cells 2.2.1. One Liter Schneider Culture Medium
1. Solution A: NaCl, Na2HPO4·2H2O, KH2PO4, KCl, MgSO4·7H2O, b-ketoglutaric acid, succinic acid, fumaric acid, malic acid. 2. Solution B: glucose, trehalose. 3. Solution C: b-alanine, l-asparagine, l-aspartic acid, l-cysteine, l-glutamic acid, glycine, l-histidine, l-isoleucine, l-leucine, l-methionine, l-phenylalanine, l-proline, l-serine, l-threonine, l-tryptophane, l-valine, l-cystine, l-tyrosine. 4. Solution D: Yeastolate. 5. Solution E: CaCl2. 6. Spectra/Por® Cellulose Ester Membranes.
2.2.2. Normal and Heavy Isotope-Enriched Amino Acids for SILAC Metabolic Labeling
1. l-lysine (H2N(CH2)4CH(NH2)CO2H·HCl, Sigma) (light, K0). 2. l-arginine hydrochloride (C6H14N4O2·HCl, Sigma) (light, R0). 3. l-Lysine-13C6, 15N2 hydrochloride (H215N(13CH2)413CH(15N H2)13CO2H·HCl, Sigma) (heavy, K8). 4. l -Arginine13C6, 15hydrochloride (H215N13C(15NH)15NH(13C H2)313CH(15NH2)13COH HCl, Sigma) (heavy, R10) (see Note 1).
2.2.3. Assembly of Schneider Media
1. Prepare Solution A as follows: Dissolve in 250 ml ddH2O: NaCl
2.1 g
Na2HPO4·2H2O
0.43 g
KH2PO4
0.68 g
KCl
1.6 g
MgSO4·7H2O
3.7 g
a-ketoglutaric acid
0.2 g
Succinic acid
0.10 g
Fumaric acid
0.10 g
Malic acid
0.10 g
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
63
2. Prepare Solution B dissolving 2.0 g glucose and 2.0 g Trehalose in 50 ml ddH2O. 3. Prepare Solution C as follows (a) Dissolve in 450 ml ddH2O: b-alanine
0.55 g
l-asparagine
0.04 g
l-aspartic
0.44 g
acid
l-cysteine l-glutamic
0.07 g acid
0.88 g
Glycine
0.27 g
l-histidine
0.44 g
l-isoleucine
0.16 g
l-leucine
0.16 g
l-methionine
0.88 g
l-phenylalanine
0.16 g
l-proline
1.87 g
l-serine
0.27 g
l-threonine
0.38 g
l-tryptophane
0.11 g
L-valine
0.33 g
(b) Dissolve 0.10 g l-cystine in 50 ml hot acidified water (pH 2, with HCl, e.g., 5 ml 1N HCl/45 ml ddH2O). (c) Dissolve 0.50 g l-tyrosine in 50 ml alkaline water (pH 9 with NaOH, e.g., 5 ml 1N NaHO/45 ml ddH2O). (d) Add the cystine solution (b) and the tyrosine solution (c) slowly (drop-wise) to the general amino acid solution (a) (see Note 2). 4. Prepare Solution D dissolving 2.0 g Yeastolate in 50 ml ddH2O. Dialyze over night against 5 L 0.9% NaCl in MCWO 3,500 Da tubes (see Note 3). 5. Prepare Solution E dissolving 0.60 g CaCl2 in 50 ml ddH2O. 6. Assemble the “minimal Schneider medium” as follows: (a) Combine sequentially solutions A–E. (b) Adjust slowly pH to 6.7 with 1N KOH (~15 ml). (c) Bring the titrated A–E mix to a final volume of 1 L. (d) Sterilize by filtration and keep at 4°C.
64
Cuomo and Bonaldi
7. Stock solution (100×) of the labeling amino acids (either light or heavy): l-arginine
40 g/L in PBS
l-lysine
165 g/L inPBS
HCl
8. “Complete SILAC-Schneider Medium” (500 ml) Combine: (a) 440 ml “minimal Schneider medium”. (b) 5 ml arginine/lysine (either light or heavy) stock (100×). (c) 3 ml Glutamine. (d) 5 ml Pennicilline/Streptomycine (100×). (e) 50 ml Serum (dialyzed against 0.9% NaCl in MWCO 3,500 Da tubes) (see Note 3). (f) Filter-sterilize the SILAC medium and store at 4°C for up to 3 months. 2.3. RNAi Interference of SL2 Cells 2.3.1. Double-Stranded dsRNA Preparation
1. Primers for PCR amplification of target gene, containing a 5′ T7 RNA Pol binding site, followed by sequences specific for the target gene. 2. High Pure PCR Purification Kit (Roche, Molecular Biochemicals). 3. MEGASCRIPT T7 transcription kit (Ambion, Austin, TX). 4. PCR Thermocycler.
2.3.2. RNAi Interference Experiment
1. 5–10 × 106 cells seeded in 10 cm culture dishes.
2.4. Sample Preparation
1. Swelling buffer: 20 mM HEPES-KOH pH 7.5, 250 mM Sucrose, 0.5 mM EDTA 1 mM DTT. (Protease Inhibitor Cocktail Tablets from Roche Applied Science to be added just before use).
2.4.1. Cells Extraction and Fractionation
2. 15–30 µg dsRNA.
2. PBS: 2 g/L NaCl, 2 g/L KH2PO4, 11.5 g/L Na2HPO2. 3. RIPA buffer: 50 mM Tris-HCl pH 7.5 150 mM NaCl, 1% NP-40, 0.5% deoxycholate, 0.1% SDS, 0.4 mM EDTA, 10% glycerol (see Note 4). 4. TCA: trichloroacetic acid. 5. Cell-homogenizer (Isobiotec, Heidelberg, Germany). 6. Vacufuge Concentrator 5301.
2.4.2. SDS-PAGE for Protein Separation
1. Precast 4/12% Tris-HCl polyacrylamide gel: NuPAGE® Novex Bis-Tris gels (Invitrogen). 2. NuPAGE® LDS Sample Buffer (4×) (Invitrogen) (see Note 5). 3. Colloidal Blue Staining Kit (Invitrogen).
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics 2.4.3. In-Gel Digestion
65
1. Acetonitrile (ACN) (HPLC grade). 2. Ammonium hydrogen carbonate (ABC) (Sigma). 3. Trifluoroacetic Acid (TFA) (Sigma). 4. Milli-Q® Ultrapure Water Purification Systems. 5. Digestion buffer: 50 mM ABC in water (pH 8.0). 6. Distaining buffer: 25 mM ABC/50% ACN. 7. Trypsin solution: 12.5 ng/mL sequencing grade trypsin (Promega Corporation, Madison, WI) in 50 mM ABC (see Note 6). 8. Extraction buffer: 3% TFA/30% ACN. 9. StageTips (18): Empore SPE C18 disks (3 M).
2.5. High Performance Liquid Chromatography– Tandem Mass Spectrometry
1. Acetic Acid (AA). 2. HPLC solvent “A”: 0.5% AA in water. 3. HPLC solvent “B”: 0.5% AA/100% ACN in water. 4. Reversed-phase material for nano-HPLC column: ReprosilPur C18-AQ, 3 mm (Dr. Maisch). 5. Agilent 1100 Series (Agilent Technologies), comprising a solvent degasser, a nanoflow pump, and a thermostated microautosampler. 6. Silica transfer line 20-cm long, 25-mm inner diameter (Composite Metals). 7. Micro Tee-connector (Upchurch). 8. LTQ orbitrap mass spectrometer (Thermo Fisher Scientific). 9. Ion Source Kit for Thermo LTQ-FT (Proxeon Biosystems).
2.6. Data Acquisition and Analysis
1. Thermo Fisher Scientific Xcalibur software 2.0.5. 2. RAWMSM 2.1. 3. MaxQuant 1.0.13.13. 4. MASCOT 2.2. 5. PC with 2 GB of RAM, dual core processor, Windows XP (32-bit version).
3. Methods 3.1. Establishment of SILAC in Drosophila SL2 Cells
SILAC has originally been developed using mammalian cells, where full incorporation can easily be achieved using isotopecoded essential amino acids in the media. Normally, Drosophila cells are grown in Schneider medium, differing from mammalian media for the presence and concentration of various components. Schneider medium is not purchased in a formula adapted for
66
Cuomo and Bonaldi
SILAC labeling (e.g., specifically depleted of the amino acids arginine and lysine), thus we assemble it from each component, based on the published receipt (16, 17). Furthermore, serum and total yeast extract, supplemented to the broth during culture, can be both source of free amino acids interfering with labeling efficiency. As such, we introduce a specific step of dialysis (see Notes 3 and 7). Cell growth and viability in SILAC medium are carefully compared with those of cells growing in standard medium, to detect any alteration from physiology that might be caused by the poorer growth conditions: 1. Inspect cells at the microscope to uncover potential gross morphological alterations. 2. Count cells and plot growth curves of cells in SILAC versus standard medium. 3. Compare cell mortality by Trypan Blue staining that allows distinguishing viable from dead cells at the microscope: at each day, the percentage of viable versus dead/dying cells is calculated (see Note 8). 3.2. Cell Extraction and Subcellular Fractionation
Fractionating the total cellular extracts reduce sample complexity with consequent increase in the dynamic range of protein identification by LC-MSMS (see Notes 9 and 10). 1. Mix equal number of cells for each sample (heavy and light). 2. Wash in PBS. 3. Incubate in 4 ml swelling buffer at 4°C for 10 min. 4. Pass the cell suspension four times through a cell-homogenizer (Isobiotec) containing a ball allowing for 10 µm clearance (see Note 11). 5. Centrifuge the cell lysate at 1,500 × g for 10 min. 6. Dissolve the pellet containing the nuclei (N) in 200 µl of RIPA buffer 7. Centrifuge the supernatant at 120,000 × g for 15 min. 7. Concentrate the supernatant (cytosolic fraction, C) by TCA (15%) precipitation and directly resuspend in SDS-PAGE loading buffer. 8. Wash the Pellet (membrane fraction, M) with PBS and resuspend in RIPA buffer.
3.3. SDS-PAGE and In-Gel Digestion
1. Resolve the proteins mixture in polyacrylamide gel. In this protocol, protein separation was achieved on a 4–12% gradient NuPAGE Novex Bis-Tris gels (Invitrogen, Carlsbad, CA) (see Note 12). 2. Cut the lanes in 10–15 slices (~5 mm broad), and each slice in small cubes, 1 mm wide.
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
67
3. Destain the gel slices with 25 mM ABC/50% ACN and add absolute ACN to shrink the gels. Repeat until the gels are completely destained. Washes are carried out in a Thermomixer, with strong shaking (>1,000 rpm) at RT. 4. Add reduction buffer (10 mM DTT) to the gel pieces; incubate for 1 h at 56°C. 5. Remove reduction buffer and add alkylation buffer (55 mM IAA); incubate for 45 min at RT, in the dark. 6. Remove alkylation buffer and repeat twice the washes (step 3). 7. Remove ACN by aspiration and dry the gels in vacuum centrifuge. 8. Rehydrate the gel pieces with ice-cold 12.5 ng/mL trypsin solution in 25 mM ABC and incubate on ice till the gels are fully swollen. Remove trypsin solution in excess. 9. Add 50 mM ABC to completely cover the gel pieces. Incubate overnight at 37°C. 10. After overnight, spin down evaporated buffer and collect liquid part in a new tube. 11. Add the extraction buffer (30% ACN, 2% TFA) to the gel pieces; incubate in a Thermomixer with strong agitation for 20 min at RT. Repeat twice. 12. Pool all supernatants. Lyophilize the peptide mixture in a vacuum centrifuge. 13. Reconstitute dried samples in 1% TFA. 14. Desalt and concentrate peptides on a reversed phase C18 microcolumn (StageTip or equivalent) as previously described (18). 15. Elute peptides from the C18 using 80% ACN/0.5% AA. 16. Remove the organic component by evaporating in a vacuum centrifuge and resuspend the peptides in a suitable injection volume (typically 5–10 µL) of 0.5% AA. 17. Inject approximately 1–5 µg of peptide into a column for nanoLC–MS analysis. 3.4. L C–MS Analysis 3.4.1. Liquid Chromatography Analysis
In this step, the peptides from in-gel digestion are separated by HPLC and introduced into MS via an on-line nanoelectrospray system. 1. Pack an analytical column in a 15-cm fused silica emitter (Proxeon Biosystems, 75-mm inner diameter), with methanol slurry of reverse-phase C18 resin at a constant helium pressure (50 bar) using a bomb-loader device (Proxeon Biosystems), as described previously (19). 2. Connect the packed emitter (C18 RP HPLC column) directly to the outlet of the 6-port valve of the HPLC autosampler
68
Cuomo and Bonaldi
through a 20-cm long (25-mm ID) fused silica without using precolumn or split device. 3. Load the tryptic peptides mixture onto C18 column at a flow of 500 nL/min. 4. After sample loading, apply a gradient of 3–60% mobile phase B at 250 nL/min over 120 min, for peptide mixture elution. 3.4.2. Mass Spectrometry Analysis
Mass spectrometry analysis is carried out on an LTQ Orbitrap mass spectrometer (Thermo Fisher Scientific) (see Note 13). 1. Operate the mass spectrometer in the data-dependent mode to automatically switch between MS and MSMS using the Xcalibur 2.4 software package. 2. Use the following settings in the “Tune” acquisition software: (a) FT full scan: accumulation target value 1 × 106; maximum filling time 1 s. (b) IT MSn: accumulation target value 10 × 103; maximum filling time 250 ms. 3. In the “Xcalibur Instrument Setup,” create a data-dependent acquisition method in which full scan MS spectra, typically in the m/z range from 300 to 1,800 amu, are acquired by the orbitrap detector with resolution R = 60,000 (see Note 14). 4. For accurate mass measurements, enable the “lock mass” option in both MS and MSMS mode (20) (see Note 15). 5. Standard acquisition method settings: (a) Electrospray voltage, 2.4 kV. (b) No sheath and auxiliary gas flow. (c) Ion transfer (heated) capillary temperature, 190°C. (d) Collision gas pressure, 1.3 mtorr. (e) Dynamic exclusion of up to 300 precursor ions for 90 s upon MSMS; exclusion mass width of 10 ppm. (f) Normalized collision energy using wide-band activation mode 35%. (g) Ion selection thresholds: 1,000 counts. (h) Activation q = 0.25; Activation time = 30 ms.
3.5. Data Analysis
All raw data files acquired are analyzed with in-house developed quantitative proteomics software MaxQuant, version 1.0.13.13 (12). MaxQuant is designed to analyze large and high-resolution MS data sets. While for the detailed description of rationale and algorithms, we advise reading the dedicated publication (12), we refer to Note 16 for a synthetic description of the modules composing this analytical platform.
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
69
The data are processed according to the following parameters: 1. Mascot search: (a) Fixed modification: Carbamidomethylation (see Note 17). (b) Variable modification: N-acetyl (Protein), Oxidation (M). (c) Missed cleavages up to 3. (d) Mass-accuracy of the parent ions in the initial Mascot search: 7 ppm. (e) Mass-accuracy for MSMS peaks in the initial Mascot search: 0.5 Da. 2. Parameters for MaxQuant: (a) Peptide false discovery rates (FDR) (see Note 18): 0.01. (b) Protein false discovery rates (FDR) (see Note 18): 0.01. (c) Maximum posterior error probability (12) (see Note 19): 1. (d) Minimum peptide length: 5. (e) Minimum number of peptides: 2. (f) Minimum number of unique peptides: 1. 3.6. Incorporation Test (see Note 20)
Usually cells are cultured in heavy medium for at least six cell doublings to allow the incorporation of the heavy amino acids. The incorporation level is measured by the proportion of remnant light peptides found in the heavy-labeled cell sample. 1. Grow cells in “heavy” medium at 26°C for at least six replications. 2. Resolve whole cell extracts from heavy-labeled cells by SDS/ PAGE. 3. Follow the protocol for in-gel digestion and subsequent MS analysis, as described (Subheadings 3.3 and 3.4). 4. Ideally, peptides identified from this pool should contain only heavy amino acids, without detectable signal at m/z values corresponding to the light peptide; however, in reality, light peptides are remaining and nonnormalized ratio H/L is therefore measured (see Notes 21–23).
3.7. One-to-One Mixture of Differentially Labeled Cells (see Note 24)
For the acquisition of a “one-to-one Schneider SL2 proteome”: 1. Grow SL2 cells in both light (L) and heavy (H) media, for the optimal number of replications indicated by the incorporation test (in this case about six replications). 2. Harvest the cells and mix in equal amount. 3. Resolve the protein extract in SDS-PAGE. 4. Digest proteins with Trypsin, as described (Subheading 3.3).
70
Cuomo and Bonaldi
5. Analyze the peptide mixture by mass spectrometry. Plot the histogram of normalized log2(H/L) ratios for the oneto-one SILAC proteome (Fig. 2a) (see Note 25). 3.8. RNAi Interference in SILAC-Labeled SL2 Cells 3.8.1. dsRNA Production
1. Amplify by PCR the individual DNA fragments, approximately 700 bp long and containing coding sequences for the proteins to be silenced. Each primer used in the PCR contained a 5¢ T7 RNA polymerase binding site, followed by sequences specific for the targeted genes. 2. Purify the PCR products with the High Pure PCR Purification Kit (Roche). 3. Produce dsRNA from the purified PCR products using a MEGASCRIPT T7 transcription kit (Ambion). 4. Ethanol-precipitate the dsRNA products and redissolve in water. 5. Reannealing can be improved by incubation at 65°C for 30 min, followed by slow cooling to room temperature (see Note 26).
3.8.2. RNA Interference of SILAC-Labeled SL2 Cells (see Note 27)
1. Seed ±8 × 106 SL2 cells in 10 cm dishes in custom serum-free medium w/o serum but w/ amino acid isotopes. 2. Add ±25 µg dsRNA to the cells and incubate for 1 h at 26°C. 3. Add serum-containing heavy or light media (equal volume) and let cells grow till harvest (see Note 28).
3.9. Experimental Set-Up for the SILAC/ RNAi Combined Approach
1. The experimental set-up for a hypothetical experiment where a specific gene product (in this case the chromatin remodeling factor ISWI) is depleted by RNAi in SL2 cells differentially SILAC-labeled is schematized in Fig. 1a. 2. In a typical “forward” labeling setting, the protein of interest is knocked-down in heavy medium, while a “mock” RNAi against GST (or any other nonendogenously expressed protein) in light medium serves as a control sample. 3. Cells are switched to SILAC medium at the time of dsRNA incubation. 4. Cells are harvested after 7 days, mixed in equal amounts. 5. Samples are fractionated in nuclear, cytosolic, membrane factions (N, C, and M respectively) (Fig. 1b). 6. Proteins are separated, digested, and analyzed by MS (Fig. 1c–e). 7. The mass spectrometry-based readout of the RNAi-SILAC experiment is schematically depicted in Fig. 1f as follows: the target protein itself is depleted from the heavy population
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
71
Fig. 1. Overview of the analytical strategy combining SILAC with RNAi (a) heavy (Arg10, Lys8) and light (Arg0, Lys0) SILAC-labeled cells are subjected to ISWI and GST RNAi, respectively. Seven days after dsRNA incubation, cells are harvested and mixed in equal amount (1:1). (b) Cells are extracted and subcellular fractionated. (c) The three fractions are loaded, in triplicate, on SDS-PAGE, and each lane is cut into ten slices. (d) Proteins are subjected to tryptic in-gel digestion, (e) peptides mixtures analyzed by nLC-MSMS on a LTQ-Orbitrap. (f) The schematized mass spectrum shows peak-pairs which corresponds to light- and heavy-peptides: significant responders were identified by peak ratios differing from one (dotted lines, circles, stars and hexagons), while nonresponders had a ratio of one (continued lines, ovals and rectangles).
relative to the light one and this should result in a peptide fold change corresponding to the knock-down efficiency (dotted lines + circles). Other significant responders should also show peak ratios differing from value 1, either increased or decreased (dotted lines, stars and hexagons). By contrast, non responding proteins should be present in similar amounts in both, heavy and light form, with a ratio of 1 (continued lines, ovals and rectangles) (see Note 29). The statistical analysis of the ratio distribution for the whole SILAC ~ RNAi proteome will lead the identification of significant responders, both up- and down-regulated upon depletion of a specific gene product (Fig. 2b, gray-dotted lines represent statistically significant outliers, 5% of the total). However, the direct comparison between the ratio distributions of the “RNAi experiment” and of the “one-to-one experiment” appears to be more informative of the global response of
Cuomo and Bonaldi
a
H/L ratio distribution (1:1 exp)
b
ratio distribution (RNAi exp)
c
Overlay distributions (RNAi vs 1:1)
−1
0
log2(H/L)
1
2
1.5
Density
1.0 0.0
0 −2
functionally up-regulated responders to RNAi
functionally down-regulated responders to RNAi
0.5
Density
1.5
statistical outliers (up)
statistical outliers (down)
0.0
1
0.5
1.0
3 2
Density
4
2.0
2.0
5
2.5
2.5
72
−2
−1
0
log2(H/L)
1
2
−2
−1
0
1
2
log2(H/L)
Fig. 2. Ratio distributions of “one-to-one” and “RNAi” experiments, and their comparison. (a) Distribution of H/L ratios in the one-to-one mixture. The histogram of log-transformed normalized H/L ratios (n = 2,449) fits a normal distribution, with standard deviation 0.12. (b) Distribution of H/L ratios in the ISWI/GST-RNAi SL2 proteome. The histogram of logtransformed normalized ratios (n = 3,976) fits a normal distribution, with standard deviation 0.19. The gray-dotted lines represent STDEV = 2 and define statistical up- and downregulated outliers (dark-gray bins, about 5% of the total) (c) Overlay of the fitted and centered H/L ratio distributions of the RNAi and one-to-one control (dotted) experiments. The overlay allows establishing the threshold of the fold change upon RNA interference, compared to the background noise measured by the 1:1. Responders that changes slightly but significantly upon the perturbation are detected and marked with indicated by light-gray bins.
the system to the gene product ablation, especially when a proteome-wide perspective is adopted. As exemplified in Fig. 2c, the broader distribution of the ISWI-RNAi experiment compared to that of the one-to-one control suggests that the cells respond with a slight but significant abundance modulation of a large number of proteins, in both directions (up- and down-regulation, light-gray bins) (see Note 30). A plain statistical analysis would miss these responders, because it identifies only the “extreme outliers”, frequently not representative of the global response of a system.
4. Notes 1. Heavy isotope amino acids used in SILAC are not radioactive and thus do not require special handling precautions. 2. The final volume of solution C should be 550 ml; however, it is advisable to limit the volume as much as possible, to avoid the risk of exceeding the maximum volume during the pH titration step. 3. Different dialysis protocols need to be tested to find the optimal compromise between the removal of free amino acids and the efficient growth of cells. In this case, about 20 h of dialysis using a MW-cut off (MCWO) of 3,500 Da for serum and yeastolate guarantee both cell growth and efficient heavy amino acids incorporation.
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
73
4. Prepare as a stock solution and store at room temperature (25°C). Fresh protease inhibitors should be added just before use. Here we use the Protease Inhibitor Cocktail Tablets Complete® (Roche Applied Science). 5. Add DTT to a final concentration of 100 mM, just before boiling. 6. Trypsin stock solution can be stored at −80°C for 6 months. 7. In some cases, more sensitive cell types might suffer from the use of dialyzed serum or other components, due to the loss of small peptides working as growth factors. If so, supplementing the SILAC medium with single purified growth factors or with a small percentage of normal serum might compensate this. 8. Generating a detailed growth curve is a useful tool to evaluate the growth characteristics of a cell line. From a growth curve, the lag time, population doubling time, and saturation density can be determined. Plot the cell number on a log-linear scale: the population-doubling time can be determined by identifying a cell number along the exponential phase of the curve, tracing the curve until that number has doubled, and calculating the time between the two. In our case, cell growth is slightly reduced in SILAC, when compared to normal medium. However, when cells are cultured in SILAC medium for prolonged periods, up to 2 months, a detailed investigation revealed neither reduction in viability nor morphological alterations. 9. The full characterization of a given proteome by LC/MSMS analysis can be limited by three instrumental factors: sensitivity, sequencing speed, and dynamic range (21). Sensitivity can be a limiting factor, depending on the starting amount of sample: when the amount is low, sensitivity decreases accordingly, in any case. Another major limitation is due to the “scan time” of the mass analyzer, which is relatively long (~1 s) and thus might not be competitive with the elution time scale of each peptides during chromatographic separation, with a consequent decrease in the sequencing capacity of the instrument and in the identification rate of peptides. Finally, proteome coverage can be limited by the “dynamic range” of the instrument, in relation to the dynamic range of the proteome to be analyzed: protein copy number is distributed from 7 to 8 orders of magnitude in cells, a range that cannot be covered by a mass spectrometer. New generation hybrid mass spectrometers started to overcome these limitations, as in the case of LTQ-Orbitrap (see Note 12); however, an efficient sample prefractionation remains an excellent strategy to improve the dynamic range of LC/MSMS analysis (see Notes 10 and 11).
74
Cuomo and Bonaldi
10. Several techniques alternative to classical SDS-PAGE can be used to obtain better sample prefractionation. In our experience, a recently introduced strategy for the separation of digested peptide from total extracts by isoelectric focusing along immobilized IPGs, named OFFGEL (Agilent), represents an attractive method for high throughput analysis; in fact, it demonstrated optimal performance, combining a higher number of protein identified by MS with diminished work-up requirements (15, 22). 11. Subcellular fractionation can be efficiently achieved also by means of alternative fractionation protocols, including kits from different brands. However, fractionation should be optimized to the type of cells employed, before setting up the SILAC experiment. 12. Reduce the risk of keratin contaminations that could interfere with LC/MS by following the procedures for sample preparation previously described (23, 24). 13. The availability of new generation of high-performance hybrid mass spectrometers such as LTQ Orbitrap allows an average absolute measurement mass accuracy (MMA) less than one parts-per-million (ppm). This is of utmost importance for two main reasons: first, high MMA of precursor ions enable using very low ppm tolerance in database search, which increases identification confidence; second, high MMA allows high-resolution MS, leading to a better peak definition with a consequent positive effect on quantitation in full MS. 14. In our case, for better sampling of the higher m/z ions, the LTQ-Orbitrap mass spectrometer operate in a “2-range” regime: an acquisition cycle consisted of two survey scans in the Orbitrap analyzer (mass ranges of 300–1,000, and 950– 1,650, at resolution R = 60,000), each followed by MSMS of the five most intense ions in the LTQ. 15. Use the background polydimethylcyclosiloxane (PCM) ions generated from ambient air (e.g., m/z = 445.120025) for internal recalibration in real time. If the fragment ion measurements are performed in the Orbitrap, use the PCM ion at m/z = 429.088735 (PCM with neutral methane loss). 16. MaxQuant works in combination with Mascot engine software (25) and is composed of two modules: “Quant.exe” assembles isotope patterns into SILAC pairs before MSMS data submission to Mascot. This model includes a sophisticated three-dimensional peak and isotope pattern detection. The output files consist of a “peak list” (.msm) of MSMS spectra, also including information relative to the Liquid Chromatography/full MS runs, such as retention times, peak intensity, etc. The “.msm files” produced by “Quant.exe” are subjected to Mascot search engine; “Identify.exe” module
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
75
takes the search results, the raw files as well as all results from the “Quant.exe” and performs robust statistical validation, to obtain an accurate assembling of peptides into proteins and, finally, to quantify proteins. Ratios for proteins are determined as the median over all measured peptide ratios for a given protein. The significance of protein ratios is measured as: (a) Significance-A, calculated by first estimating the variance of the distribution of all protein ratios in a robust way, and then reporting the error function for the z-score corresponding to the given ratio. (b) Significance-B, which uses the same strategy, but also takes into account the dependence of the distribution on the summed protein intensity. 17. The modifications corresponding to arginine and lysine labeled with heavy stable isotopes (arginine 13C615N4 and 13 C615N2) can be treated as fixed modifications in the Mascot search, if applicable, after identification of SILAC pairs in the “Quant.exe” module. 18. False positive rates for peptides are calculated as described in (26). In this case, we fixed 1% false discovery rate, both at the level of peptides and proteins. 19. PEP is the probability of a false hit, given the peptide identification score and length of peptides (12). 20. Quantification errors will arise if proteins and peptides are not fully labeled with heavy SILAC amino acids at the end of the adaptation phase. Therefore, an analysis to check the degree of incorporation of the heavy SILAC amino acid is critical when working with a new cell line or medium formulation. 21. In practice, after 7 days of SL2 growth, the average of heavy to light (H/L) nonnormalized peptide ratios is about 30, corresponding to more than 95% incorporation efficiency (1-1/AV Ratio). 22. It is also possible to use an in-solution digestion of whole cell extract from heavy sample as previously described (27): in fact, a complete proteome is not needed to check the incorporation level and a statistically significant subset is enough. 23. In some cell types grown in standard media, the metabolic interconversion between arginine and proline can occur when arginine is provided to cells in excess. The reverse metabolic conversion of Pro-to-Arg can also occur when cells are not provided with enough arginine. As such, when assembling a new SILAC medium, the optimum concentration of arginine must be determined experimentally for the cell line under investigation. Titration of arginine in SILAC Schneider medium is followed by measuring the frequency of heavy proline in heavy SILAC-growing cells. An interesting alternative
76
Cuomo and Bonaldi
is also increasing the concentration of light Proline in the medium, to counteract Arg conversion. 24. SILAC-based quantitative proteomics relies on the assumption that the abundance of specific proteins is not affected by the isotopic composition of the SILAC medium employed. The distribution of protein ratios in the quantitative proteome from the one-to-one “light” and “heavy” mixture allows validating this hypothesis. The normal distribution of the histogram informs about the homogeneity of labeled amino acids incorporation and about any potential effect of the isotopic composition of the medium on protein levels. Furthermore, the width of the distribution correlates with the precision of the SILAC-ratio measurement for identified and quantified proteins, over the background noise. The background noise results from several factors: the biological variability among cell cultures, the random variation of protein turn-over, the precision of the instrument and others. As such, the normalized H/L ratio distribution from the oneto-one experiment represents an indicator of the overall sensitivity of the approach in detecting protein level changes upon perturbation, over the background noise. 25. In our case, the histogram of log-transformed normalized H/L ratios (n = 2,449) in the one-to-one mixture fits a normal distribution with a standard deviation of 0.12 (dotted line, Fig. 2a). 26. To ensure that the majority of the dsRNA exists as a single band of about 700 bp, a small aliquot of dsRNA is analyzed by agarose gel electrophoresis. 27. Drosophila cells are uniquely suited to RNAi-mediated gene knockdown (28, 29). Target genes can be depleted by simply incubating the cells with long dsRNA molecules, in serumfree medium. Long dsRNA are directly up-taken by pinocytosis by cells and subsequently they enter the interference splicing and processing pathway. Albeit to a minor extent, potential off-targets effects can be observed in fruit fly system, due to the unspecific silencing of targets by a certain dsRNA (30). Thus, biological replicates using alternative dsRNAs sequences are highly recommended. 28. Validation of the silencing of the protein of interest (ISWI) by RNAi is carried out by standard Western Blot. 29. To increase the biological significance of the analysis, different experimental replicates should be planned; a common strategy is to perform one experiment in a “forward” set-up, followed by at least one of the replicates in a “reverse” set-up, where the heavy and light media are “switched” between the real and the control experiment. The MS-based readout of the changes in protein ratio will consequently be inverted as a result.
Systems Biology “On-the-Fly”: SILAC-Based Quantitative Proteomics
77
30. Importantly, the overlay between fitted and centered H/L ratio distributions of the RNAi and the one-to-one control experiment allows establishing the threshold of the fold change to be considered for a comprehensive quantitation analysis. As such, a more comprehensive picture of the global response of a perturbed system can be achieved. Noteworthy, the strategy described allows detecting fine but significant changes at the protein level. In this scenario, we claim that this approach is useful for robust and reliable comparative analysis, as in the case of proteome versus transcriptome studies (31).
Acknowledgments The work in T.B. laboratory is supported by an Armenise-Harvard foundation career development program grant, a grant from the Associazione Italiana Ricerca sul Cancro (AIRC) (REF. # 6011), a grant from the Association of International Cancer Research (AICR) (REF. # 09-0281) and a grant from Cariplo Foundation (REF. # 2009-2721). References 1. Gygi SP, Rochon Y, Franza BR, Aebersold R (1999) Correlation between protein and mRNA abundance in yeast. Mol Cell Biol 19:1720–1730 2. Lu P, Vogel C, Wang R, Yao X, Marcotte EM (2007) Absolute protein expression profiling estimates the relative contributions of transcriptional and translational regulation. Nat Biotechnol 25:117–124 3. de Groot MJ, Daran-Lapujade P, van Breukelen B, Knijnenburg TA, de Hulster EA, Reinders MJ, Pronk JT, Heck AJ, Slijper M (2007) Quantitative proteomics and transcriptomics of anaerobic and aerobic yeast cultures reveals post-transcriptional regulation of key cellular processes. Microbiology 153:3864–3878 4. Tian Q, Stepaniants SB, Mao M, Weng L, Feetham MC, Doyle MJ, Yi EC, Dai H, Thorsson V, Eng J, Goodlett D, Berger JP, Gunter B, Linseley PS, Stoughton RB, Aebersold R, Collins SJ, Hanlon WA, Hood LE (2004) Integrated genomic and proteomic analyses of gene expression in Mammalian cells. Mol Cell Proteomics 3:960–969 5. Nie L, Wu G, Culley DE, Scholten JC, Zhang W (2007) Integrative analysis of transcriptomic and proteomic data: challenges, solutions and applications. Crit Rev Biotechnol 27:63–75
6. Tyers M, Mann M (2003) From genomics to proteomics. Nature 422:193–197 7. Dorus S, Busby SA, Gerike U, Shabanowitz J, Hunt DF, Karr TL (2006) Genomic and functional evolution of the Drosophila melanogaster sperm proteome. Nat Genet 38:1440–1445 8. Newman JR, Ghaemmaghami S, Ihmels J, Breslow DK, Noble M, DeRisi JL, Weissman JS (2006) Single-cell proteomic analysis of S. cerevisiae reveals the architecture of biological noise. Nature 441:840–846 9. Domon B, Aebersold R (2006) Mass spectrometry and protein analysis. Science 312(5771):212–217 10. Kuster B, Mann M (1998) Identifying proteins and post-translational modifications by mass spectrometry. Curr Opin Struct Biol 8:393–400 11. Nielsen ML, Savitski MM, Zubarev RA (2005) Improving protein identification using complementary fragmentation techniques in fourier transform mass spectrometry. Mol Cell Proteomics 4(6):835–845 12. Cox J, Mann M (2008) MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat Biotechnol 26:1367–1372
78
Cuomo and Bonaldi
13. Boutros M, Bras LP, Huber W (2006) Analysis of cell-based RNAi screens. Genome Biol 7:R66 14. Boutros M, Kiger AA, Armknecht S, Kerr K, Hild M, Koch B, Haas SA, Paro R, Perrimon N (2004) Genome-wide RNAi analysis of growth and viability in Drosophila cells. Science 303:832–835 15. Bonaldi T, Straub T, Cox J, Kumar C, Becker PB, Mann M (2008) Combined use of RNAi and quantitative proteomics to study gene function in Drosophila. Mol Cell 31:762–772 16. Ong SE, Mann M (2007) Stable isotope labeling by amino acids in cell culture for quantitative proteomics. Methods Mol Biol 359:37–52 17. Ong SE, Mann M (2006) A practical recipe for stable isotope labeling by amino acids in cell culture (SILAC). Nat Protoc 1:2650–2660 18. Rappsilber J, Ishihama Y, Mann M (2003) Stop and go extraction tips for matrix-assisted laser desorption/ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics. Anal Chem 75:663–670 19. Olsen JV, Ong SE, Mann M (2004) Trypsin cleaves exclusively C-terminal to arginine and lysine residues. Mol Cell Proteomics 3:608–614 20. Olsen JV, de Godoy LM, Li G, Macek B, Mortensen P, Pesch R, Makarov A, Lange O, Horning S, Mann M (2005) Parts per million mass accuracy on an Orbitrap mass spectrometer via lock mass injection into a C-trap. Mol Cell Proteomics 4:2010–2021 21. de Godoy LM, Olsen JV, de Souza GA, Li G, Mortensen P, Mann M (2006) Status of complete proteome analysis by mass spectrometry: SILAC labeled yeast as a model system. Genome Biol 7:R50 22. Hubner NC, Ren S, Mann M (2008) Peptide separation with immobilized pI strips is an attractive alternative to in-gel protein digestion for proteome analysis. Proteomics 8:4862–4872
23. Shevchenko A, Wilm M, Vorm O, Mann M (1996) Mass spectrometric sequencing of proteins silver-stained polyacrylamide gels. Anal Chem 68:850–858 24. Shevchenko A, Tomas H, Havlis J, Olsen JV, Mann M (2006) In-gel digestion for mass spectrometric characterization of proteins and proteomes. Nat Protoc 1:2856–2860 25. Perkins DN, Pappin DJ, Creasy DM, Cottrell JS (1999) Probability-based protein identification by searching sequence databases using mass spectrometry data. Electrophoresis 20:3551–3567 26. Nesvizhskii AI, Vitek O, Aebersold R (2007) Analysis and validation of proteomic data generated by tandem mass spectrometry. Nat Methods 4:787–797 27. Ong SE, Mann M (2006) Identifying and quantifying sites of protein methylation by heavy methyl SILAC. Curr Protoc Protein Sci Chapter 14, Unit 14.19 28. Armknecht S, Boutros M, Kiger A, Nybakken K, Mathey-Prevot B, Perrimon N (2005) Highthroughput RNA interference screens in Drosophila tissue culture cells. Methods Enzymol 392:55–73 29. Sachse C, Krausz E, Kronke A, Hannus M, Walsh A, Grabner A, Ovcharenko D, Dorris D, Trudel C, Sonnichsen B, Echeverri CJ (2005) High-throughput RNA interference strategies for target discovery and validation by using synthetic short interfering RNAs: functional genomics investigations of biological pathways. Methods Enzymol 392:242–277 30. Ma Y, Creanga A, Lum L, Beachy PA (2006) Prevalence of off-target effects in Drosophila RNA interference screens. Nature 443: 359–363 31. Schrimpf SP, Weiss M, Reiter L, Ahrens CH, Jovanovic M, Malmstrom J, Brunner E, Mohanty S, Lercher MJ, Hunziker PE, Aebersold R, von Mering C, Hengartner MO (2009) Comparative functional analysis of the Caenorhabditis elegans and Drosophila melanogaster proteomes. PLoS Biol 7:e48
Chapter 4 Systems Biology of Cell Behavior Najl V. Valeyev, Declan G. Bates, Yoshinori Umezawa, Antonina N. Gizatullina, and Nikolay V. Kotov Abstract Systems Biology approaches to drug discovery largely focus on the increasing understanding of intracellular and cellular circuits, by computational representation of a molecular system followed by parameter validation against experimental data. This chapter outlines a universal approach to systems biology that allows the linking of intracellular molecular machinery and cellular activity. This procedure is achieved by applying mathematical modeling to molecular modules of a cell in the light of systems biology techniques. Key words: Cell activity, Cell behavior, Paramecium, Mathematical, Modeling, Systems biology
1. Introduction The term “Systems Biology” has a very deep meaning. Essentially, it is an attempt to represent Biology as a System and to apply to it the full power of various system-specific techniques developed in physical, engineering and philosophical sciences. In drug discovery, systems biology models can be used to rank compounds or optimize treatment protocols. A number of complex diseases are due to the altered “tuning” of the system, and they can be caused by a combination of environmental and internal factors, rather than by the up- or down-regulation of individual genes. In such cases, cells would still have normal phenotypes. However, impaired signaling between cells can be sufficient to cause a disease, even if it leaves insufficient signs to easily identify diseased cells or pathways. The cause of such diseases lies at the systems level. In the cases of systems biology diseases, the treatment
Qing Yan (ed.), Systems Biology in Drug Discovery and Development: Methods and Protocols, Methods in Molecular Biology, vol. 662, DOI 10.1007/978-1-60761-800-3_4, © Springer Science+Business Media, LLC 2010
79
80
Valeyev et al.
strategies require understanding, classifying and modeling the corresponding mechanisms of cell activity. There are at least three levels where systems approaches can be utilized in pharmaceutical development: signal reception, intracellular responses, and intercellular communication between various cell type populations (1). A meaningful way of representing a biological system is to divide it into functional modules (2). A classical analysis that highlights the differences between traditional sets of tools employed in biology and engineering was published by Yuri Lazebnik (3).
2. Philosophical Foundations of Systems Biology
It has recently been recognized that robust mathematical representation of biological systems can be achieved by utilizing a number of philosophical principles (4). Any attempt to model a biological system always faces a significant degree of ambiguity, since there are always multiple possibilities for the mathematical representation of a biological system. This ambiguity can be significantly reduced by focusing on the most essential core biological processes which are identified when a cell is analyzed in the light of critical tasks it requires for its functionality and survival. Here, we propose a universal approach for developing a systems biology framework which links intracellular molecular machinery with cellular activity. Such a procedure can be achieved by applying mathematical modeling to molecular modules of a cell in the light of philosophical systems biology approaches (4, 5). Behavior in the most general sense is described in the introduction to the Tinbergen book: “If life is the most complex state of a substance, the behavior, undoubtedly, is the most complex representation of vital activity. Everything that takes place in an organism, chemical, physical and physiological processes, ultimately results in the external activity, in cell behavior” page 7 in (6). This citation is applicable to the cell behavior as well as to multicellular organisms. It has been demonstrated that the behavior of single cell organisms as well as that of individual cells of multicellular organisms can be very complex (7). Currently, a large number of methodologies have been developed to analyze complex systems in different areas of human activity. For example, engineering deals with systems of arguably comparable levels of complexity as a single cell or at least a pathway. Many engineering principles have been reported to be transferable to biology, with a number of successful examples now appearing in the literature (8–10). The cell behavior mechanisms have continuously been optimizing during evolution. One way to
Systems Biology of Cell Behavior
81
understand the underlying molecular mechanisms behind cell behavior is to reconstruct the development of molecular circuits in relation to cell activity. A philosophical diagram for the systems biology representation of cell behavior is shown in Fig. 1. It is adopted from a general methodological approach of object representation as a system developed by (5). The diagram shows a number of possible representations of a cell, by “illuminating” it from a number of angles. One representation, which is the most well practiced in biology, is the morphological representation of the cell, which shows the internal protein composition as well as other “parts” connected with each other. Cell behavior can also be seen from a “substrate”-oriented point of view. Another efficient representation of a cell is as a combination of functional modules. The next level of analysis implies the existence of signaling processes between the connected parts. The multitude of representations shown in Fig. 1 is in many cases already well developed in various areas of engineering, but still insufficiently recognized in biology. It is one of the reasons why most faulty objects ranging from a simple radio to a modern space shuttle can always be fixed by a qualified engineer. The same cannot be said about people with disease. Each projection of a cell representation needs to be analyzed under idealized conditions. Such idealized representations have a long history in science. For example, Galileo introduced
Fig. 1. Functional decomposition of a living cell. A cell is highlighted from a number of angles that can reveal core fundamental processes defining cell behavior. Such decomposition is performed in the morphological and idealized representations. The outlined range of cellular “views” provides guidance for the direction of mathematical model development that would link the molecular mechanism with physiological phenotype.
82
Valeyev et al.
basic principles in physics by stating that any object would move continuously and indefinitely with the same speed, if it undergoes such motion under idealized conditions with no friction and no influences from other objects. According to Plato, each engineered system has a backbone idea, which can be converted and represented as functional and procedural diagrams (11) relating the “parts” of the system to each other. The connectivity of those parts would create a system satisfying a predefined set of properties that include the requirements of reliability, immunity, consistency, and criteria of quality. Biological evolution from an engineering point of view can be seen as a development of a system that would allow the organism’s survival in a given ecological niche. The Systems Biology approach is based on this backbone idea. Understanding biology from a philosophical system’s point of view implies reconstructing the backbone idea of evolutionary engineered systems on a given biological material, that is, proteins, genes and other cell “components”. For example, the backbone idea of television is the representation of pictures and sounds as linear sequence of signals, translation of the signals over a distance followed by the picture and sound reconstruction. The backbone idea in the Paramecium system example proposed in this chapter is the survival of species mediated by cilium-dependent movements in a complex heterogeneous environment, by feeding on small organisms, organizing reproduction, and active defense from predators while at the same time remaining a single cell organism. This idea is realized in the cell morphology with unique reception and effector systems creating the molecular machine organizing the overall cell behavior. The backbone processes found in any living organism include reproduction, defense, evolutionary development, and functional maintenance. First consider cell behavior in the light of the idealized backbone processes projection as shown in Fig. 1. As mentioned, reproduction can be considered as one of the basic “core” processes for any living organism. The necessity of reproduction is one of the key features that make living systems different from any other physical objects or human engineered machines. Most of the biological parts, proteins, genes, etc., are unreliable. Proteins can easily misfold, due to a mutation or a physical damage, whereas many stretches of DNA sequence can be highly polymorphic. Reproduction or replacement allows the creation of reliable biological systems composed of unreliable parts. Reproduction is required on a number of levels: reproduction of populations of species, individual organisms, tissues, cells, intracellular substructures, forming cells. On the cellular level, this general reproduction process can be further divided into a number of subprocesses: cell division, cell growth and development (protein synthesis and assembly, cell type differentiation), substrate and
Systems Biology of Cell Behavior
83
energy supply, preprogrammed cell death, the disassembly of proteins as well as other cell parts. The mechanism of reproduction is governed by a number of standard cellular activity “programs”, determined by the specifics of the ecological niche and the cell itself. An excellent example for such analysis can be drawn from the behavior of the unicellular organism Paramecium caudatum, since both the intracellular morphological composition and cell behavior under various physiological conditions have been extensively elucidated. The relatively “simple” intracellular machinery is a more “convenient” subject to study in comparison with more complex cells from multicellular organisms. Single cell organisms are also an excellent model for drug screening studies. For example, Paramecium has been used as an eukaryotic model system to study cellular effects of neuroactive drugs (12) and investigate cytotoxicity of organic solvents and organophosphorus insecticides (13).
3. Phenomenological Analysis of Paramecium Behavioral Reactions
The fact that Paramecium behavior is a combination of a limited number of behavioral reactions was observed in the beginning of the nineteenth century (14). The reported reactions include: avoidance reaction, the escape reaction, thigmotaxis (full halt of movement), trichocyst release, conjugation, chemotaxis, galvanotaxis, gravitaxis, and cell division (15). Eight more Paramecium behavioral reactions were identified recently: unlimited backward movement, the reverse motion by undergoing the cell shape folding (under conditions of limited space), the “search” reaction, which is based on numerous avoidance reaction initiations, randomized movements, the non-spiraling movement with the oral grove oriented toward the bacteria concentration, flash-dependent movement initiation from thigmotaxis reaction, local trichocyst release, and the directed movement. Consider the Paramecium behavior main underlying process related to the substrate and energy supply in the context of a functional decomposition, as shown in Fig. 1. Paramecia feed on any microorganisms with size smaller than its grove. In real conditions, the feeding substrate for Paramecium can be distributed very unevenly, for example localized either on the surface or on the bottom, in the vicinity of decaying flora remains. Therefore, the mechanism of feeding associated behavior needs to contain the following functions shown in Fig. 2: (1) search and identification of substrate (Fig. 2a), (2) dynamic and static maintenance in the area of substrate concentration (Fig. 2b), (3) optimization of grove orientation relative to the substrate presence (Fig. 2c),
84
Valeyev et al.
Fig. 2. (a) Paramecium “feeding”-associated behavioral movement phenotypes. (b) When paramecia identify favorable conditions, Paramecia switch on dynamic behavioral models to remain in the favorable zone. (c) Paramecia can also perform special type of movement to optimize the grove orientation to intake maximum amount of bacteria. (d) When conditions become unfavorable, cells can move for very significant periods of time without any substrate intake.
(4) food intake, and (5) initiation of the search process upon substrate completion (Fig. 2d). Under favorable conditions, Paramecium can also attach itself to an available surface. The second core process is the evolution of species, organisms, tissues, cells, and cell structure maintaining parts. The mechanisms of this process include mutations, sexual behavior, and genetic exchange between different species, selection, and survival of the fittest. Cells that have sexual behavior possess a number of activity programs allowing the genetic material exchange. The sexual activity of Paramecium can be schematically represented by the following scheme (Fig. 3). Cells which are prepared to conjugate release special substances to the cell surface that make the surface adhesive to a complementary clone and form a furrow without cilia. The furrows allow cellular membranes to come into close proximity. Consider the conjugation process in the context of the functional diagram shown in Fig. 1. The required procedure is the search for a sexual partner and the contact with complementary zones.
Systems Biology of Cell Behavior
85
Fig. 3. Schematic representation of the conjugation process in Paramecia.
The analysis of the Paramecium motion at the initial stages of conjugation does not reveal any specific motion patterns. The first contact appears to be a random event. Next, sexually prepared cells with adhesive surfaces start random motion with respect to each other. Figure 4 shows the experimental recordings of the Paramecia movements during the conjugation. Initial attachment of multiple cells (Fig. 4a) is followed by a separation of individual cell pairs attached to each other in a randomly oriented fashion. Once separated from the overall group of glued cells, the pair of Paramecia starts performing multiple avoidance reaction by employing abrupt cilia movements causing random mutual shift (Fig. 4b). The random continuous shift of the respective cellular positions occurs until the complementary zones come into contact to exchange genetic material. Another essential process is defense. Any organism remains alive as long as it is protected. Protection includes the behavioral defense, immunity, and the protection of DNA. Figure 5 shows the Paramecium response to danger. Figure 5a shows the Paramecium speed in response to contact with a needle. In cases when this reaction is randomly initiated (Fig. 5b) the cellular speed alteration (Fig. 5c) appears to be similar to the contact by a needle. It was also found that paramecia have a light-sensitive reaction. A flash of light causes stationary cells (undergoing the thigmotaxis reaction) to accelerate (16). 3.1. Avoidance Reaction
Avoidance reaction was one of the first observed Paramecium movement reactions and was first described in (14). As a result of the avoidance reaction, the cell changes a direction of movement, and continues movement at an angle to the previous motion path trajectory anywhere in the range from 0 to 180°. This reaction is initiated in situations when cells need to overcome mechanical barriers, in the initial stages of the conjugation process and also in order to remain in the “optimum” zone with favorable conditions.
86
Valeyev et al.
Fig. 4. (a) Randomly “glued” cells observed at initial stages of conjugation before individual pairs of cells single out from the overall “tangle”. (b) Frame by frame images of the mutual cellular movements during the conjugation process. Two cells “glued” together perform multiple avoidance reactions that lead to the change of the body orientation. The process continues until complementary zones come into contact and the exchange of genetic material takes place. (c) The cellular trajectory formed as a result of multiple execution of the avoidance reaction. (d) The relative movement (d) of two cells as a result of single avoidance reaction. (e) Oscillations of the transmembrane potential on the Paramecium membrane during the initial stages of conjugation.
3.2. Escape Reaction
In response to hyperpolarizing irritation, which can be very short, cells respond by a dramatic speed increase followed by gradual reduction to the original level. Figure 5 shows the cell speed dependence on time during the escape reaction. Electrophysiological studies show that after doubling, speed cells return to the original moving mode in about 0.4 s (17, 18).
3.3. Thigmotaxis
In situations where bacterial substrate is localized on the bottom or any other local volume in the medium, Paramecia have been shown to come to a complete halt of movement. For the first time, this reaction was described in (14, 19). In thigmotaxis, most of the cell surface cilia move with a low frequency, whereas the
Systems Biology of Cell Behavior
87
Fig. 5. (a) Escape reaction. (b) The cellular speed as a function of time in response to an irritation caused by a needle. (c) The trajectory is shown during randomly initiated escape reaction with the corresponding cellular speed.
cilia distributed around the oral grove continue intensive strokes. According to Dogel (20), the only possibility for a cell to remain completely motionless is by being attached to a surface by locally released trichocysts.
88
Valeyev et al.
3.4. Conjugation
Under certain conditions, Paramecia cells create cell couples with a bridge connecting between cytoplasmic membranes, through which the genetic material exchange occurs (21).
3.5. Chemotaxis
Paramecium cells cultivated in an experimental tube gather in the upper layer of the medium, where the oxygen and substrate levels are optimal. Cellular presence in the upper layer has an active dynamic mechanism. Cells can stay attached to the tube surface for some time, but once they start moving, the initiated avoidance reaction returns them into the area with favorable conditions. Paramecia are also sensitive to a number of chemical compounds that alter the parameters of their motion (22).
3.6. Galvanotaxis
Paramecia have been shown to move along spiral trajectories in applied electromagnetic field, with the spiral trajectory axes directed in parallel with the electromagnetic field (23–25). In the described cases, Paramecia move in the cathode direction. Paramecia have also been reported to move at an angle to the electric field or in circles.
3.7. Trichocyst Release
Paramecium body is covered by special organelles that can release thread-like fibers, in situations where cells are being attacked by a predator or in order to create an attachment to a surface.
3.8. Cell Division
Paramecium caudatum reproduces by division. It is clear that the cell is highly asymmetrical, in front it has an oral grove, and an organelle that releases substances at the rear. Shortly after division (in about 10–15 min) the two asymmetrical halves reconstruct the remaining part of the cell.
3.9. Behavior in a Dendrite
Observations of Paramecium movement trajectories reveal eight additional behavioral programs. In situations when cells find themselves in a channel type dendrite cavity with a diameter smaller than the cell length, it becomes nearly impossible to turn around by alternating the direction and frequency of cilia beat only. There are two potential possibilities in those situations; one is to start backward movement, another to turn by folding the body. Both types of movement appear to be present in Paramecia.
3.10. Stochastic Ball Movement
A special type of Paramecium movement constitutes a slow motion with constantly changing parameters of trajectory, such as spiral radius, spiral step, speed and direction of rotation with periodic spontaneous initiation of the escape reaction. In this case, trajectories are almost “entangled” in a ball. The cell does not change its location over a significant period of time. This behavior is observed in very well adapted cells in cultural medium under favorable conditions.
Systems Biology of Cell Behavior
89
3.11. Movement in the Vicinity of a Limiting Surface
Paramecium has a specific type of motion in the vicinity of a surface. While contacting a surface, the cell is moving relatively slowly without usual body rotation, constantly contacting the surface with the front part of the body. This type of movement is observed when a thin layer of bacteria is evenly distributed on the bottom or on the top area of an experimental tube.
3.12. The Search Reaction
At initial stages of conjugation, Paramecia form an aggregate from randomly glued up cells. After a period of time, pairs of cells single out from others with a strict body alignment. Analysis of the cell movements during the conjugation process shows that cells make abrupt movements as a result of abrupt cilia strokes as observed during the avoidance reaction. Cells thereby keep adjusting the mutual orientation until creating a contact by complementary zones and forming a bridge between the cellular cytoplasms. If the cell accidentally becomes unattached from the cellular aggregate, it moves in a trajectory that can be observed during the constant initiation of avoidance reaction.
3.13. The “Going Away” Reaction
Under unfavorable positions, with absent bacterial substrate or low KCL concentration, Paramecia follow a spiral trajectory with a very small radius and high speed (up to 2 mm per second). Under such speeding mode, the cellular trajectory represents almost a straight line. A cell moving with an average speed 1 mm per second can swim more than 80 m in 24 h. The substrate, accumulated by cell, can last for several days of movement. By executing this type of movement, Paramecium attempts to cover large distances in order to leave the zone with unfavorable conditions.
4. Systems Modeling of Cell Movement
One of the main advantages for systems analysis of Paramecium behavior is the relatively well-established intracellular pathways. The two key players for cilia beat regulation are the intracilia Ca2+ concentration and the transmembrane potential. While there are a number of channels, notably Na+, K+, and Ca2+ channels discovered in the cilia membrane of Paramecium, Ca2+ regulated Ca2+ channels introduce the major contribution into the generation of Ca2+ spikes that regulate the frequency and direction of cilia beat in Paramecium (26). These Ca2+-dependent Ca2+ channels are also found on the endoplasmic reticulum membrane (27, 28) and are considered as important regulators of intracellular signaling in mammalian cells (29, 30).
90
Valeyev et al.
The Ca2+ regulatory system found in Paramecium appears to be a highly conserved regulatory system in many multicellular organisms including humans. Multiple Ca2+-dependent pathways are targets for compounds being investigated in multicellular organisms (31–35). A number of mathematical models have been developed to describe the core mechanisms regulating Paramecium behavioral phenotypes by employing the philosophical systems biology principles outlined in this chapter. 4.1. Transmembrane Potential and Ca2+ Regulation Model
For the first time, a mathematical model for Ca2+ and membrane potential in Paramecium regulation was proposed in (36). In the most general case, the dynamics of Ca2+ alterations in an individual cilium can be described as: VR ·
d Ca 2 + dt
=
SR
(
P A T u + I Ca + I Ca + I Ca · I Ca
z ·F
2+
2+
2+
2+
)+ J ([Ca
2+
], · [CaM])
(4.1) where V R is cilium volume, S R is the surface area of a cilium, A P T 2+ I Ca I Ca I Ca currents due to passive Ca2+ 2+ and 2+ 2+ are the Ca ,
transport, active Ca2+ transport, and Ca2+ leak from cilium into u the cell body, respectively. I Ca2+ is the current generated by Ca2+ leakage,
(
)
A 2+ . J Ca , [CaM0] describes Ca2+ binding and m2
release by calmodulin, major Ca2+ binding regulatory protein in Paramecium cilia. z = 2 is the charge of Ca2+ ions. F is Faraday’s constant. 4.2. Intracilia Ca2+ ConcentrationDependent Paramecium Movement
The simplified equations for the Ca2+ concentration (u) and transmembrane potential (y) alterations are given by: C 00 + dψ = g + (CaM 0,3 ) ψ − ln Na in Na dτ CNa + b
u 1 + g 2+ (CaM 0,3 ) ψ − ln 0 − g + (CaM 4 ) ψ −ψ + − gl (ψ −ψ l )+ I 0 , K Ca 2 u K
(
)
1 u u du 1 u = g 2+ CaM 0,3 ψ − ln 0 + γ (CaM 4 ) − α exp (− β ψ ) ζ − ln + λ KA +u dτ Ca 2 u 2 uT
(
)
4
∑ fi (CaMi ) i=0
(4.2) z·F 2 ·V ·K Ca2+ j ·F Ca 2+ ,y = M ,b= . is the where τ = t ·k−1 , u = R·T K Ca2+ C ·S M ·R·T capacity of the membrane, S M is the membrane surface, V is the cilium volume, K Ca2+ is the maximum dissociation constant for the Ca2+–CaM binding reaction. k−1 is the Ca2+ ion dissociation constant from CaM, j M is membrane potential. y K + and y l are
Systems Biology of Cell Behavior
91
the nondimensional equilibrium potentials of K+ and leakage cur-
rents, respectively. g Na+ (CaMi ), g Ca2+ (CaMi ) , g K + (CaMi ) , g l
are nondimensional conductances. z is the nondimensional potential between cell body and cilium. uT is the nondimensional Ca2+ concentration in the cell body, f i (CaM ) is function that describes Ca2+ uptake by CaM. a, b, l ,g, are constants and I0 is an external nondimensional current. CaM0 + CaM3 , + CaM0 + CaM3 CaM0 + CaM3 , g 2+ (CaM0,3 ) = g 2+ · Ca Ca K 2+ + CaM0 + CaM3 Ca _ CaM 4 CaM 4 g + (CaM 4 ) = g + · , g (CaM 4 ) = g · , K K K + + CaM 4 K A + CaM 4 g
Na+
(CaM0,3 ) = g
Na+
·
K
Na+
K
where g Na , g K , g Ca +
+
2+
are the nondimensional conductances,
K Na+ , K Ca2+ , K K + , K A – are the nondimensional equilibrium dissociation constants for the corresponding CaM-channel complex formation reactions. The simplified description of CaM conformational dynamics is given by:
dCaMi 1 ,...CaM 4 , u ), i = 0,1 = ℑi (CaM0 ,CaM dt
.(4.3)
The analysis of the dynamic properties of the described equations reveals that the system possesses excitable properties even in the absence of any potential dependent channels. CaM conformations have distinct and selective activation properties when bound to different numbers of Ca2+ ions (37, 38). The phase diagram of coupled Eq. (2) shows that when Ca2+ and N+ channels are activated by CaM conformations with one and three Ca2+ ions, simultaneously with K+ channels and active Ca2+ transport regulation by fully Ca2+ occupied CaM, the system Eq. (2) has three dynamic states: (1) generation of a single Ca2+ spike, (2) stationary oscillations, (3) trigger. Figure 6 shows the phase diagram for the trigger type mode of the system. The alterations of intracilia Ca2+ concentrations regulates the cilia movement via changing the levels of cyclic monophosphates, such as cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate (cGMP) (39), followed by cyclic monophosphate-dependent phosphorylation of effector proteins in the base of Paramecium cilia. In Paramecium, AC is regulated by CaM bound to one Ca2+ ion, PDE for cAMP is activated by fully bound CaM, GC is governed by CaM with three Ca2+ ions,
92
Valeyev et al.
Fig. 6. Phase plot of the relationship between the Ca2+ levels and transmembrane potential. The null-clines cross at three points two of which are stable. The solution predicts the trigger-type mode of operation.
whereas the activity of PDE for cGMP is inhibited by apo CaM. The effector proteins (dynein arms) regulating cilia beat frequency are phosphorylated by both cAMP and cGMP-dependent protein kinases, while the proteins regulating the direction of effective cilia beat are phosphorylated by Ca2+-CaM-dependent protein kinase (39, 40). A mechanistic model linking the cilia beat frequency and the direction of effective strike allows calculation of cell movement trajectories as a function of intracilia Ca2+ (41). Mathematical modeling of the Ca2+ regulatory system in combination with a physical model for Paramecium cilia motion allows the mechanistic understanding of Paramecium motion under various external conditions. The combination of the philosophical approaches to systems biology, physiological observations of cell movements, the mathematical modeling of the intracilium and intracellular molecular system, and mechanical consideration of a cell allow the prediction of Paramecium reactions according to the external conditions. Figure 7 shows the modeled trajectories of Paramecium movement during the performance of individual reactions. The predicted cell movement trajectories are compared with the experimentally observed trajectories at different levels of intracilia Ca2+ concentration. Figure 7a–c show the extreme cases